The largest need, and the quiet over it

Two months ago Terafab was the whole point. The semiconductor fab was the capstone of the pitch that sold the company, the centerpiece of the June 4 interview in which Jamie Dimon of J.P. Morgan, whose bank led the offering, drew the plan out of the controller, and the controller named the fab a necessity: there would not otherwise be enough chips, and “that’s why we need to do Terafab. It seems essential” (FILED). A company puts its essential project at the front of the document that sells its shares. The fab was important by the issuer’s own hand, and it was to begin before the decade turned.

It is a build of about $119 billion, and a build that size needs its own company, its own partners, its own financing, none of which the SpaceX balance sheet supplies. On all of it, structure, money, forecast, timing, the first public quarter said nothing. Terafab was absent from the use of proceeds, absent from the quarterly, absent from the earnings release, and absent from every question on the August 4 call. Its only motion sits in a Texas county records office, where the districts approved a tax abatement and the company signed a development agreement to invest five billion dollars by 2030 against the hundred-and-nineteen-billion headline (REPORTED). On the company’s own county schedule the fab spends almost nothing before 2028. The largest number in the gap exists, in the filed and public record, as a tax form.

The build was to end the need to buy chips at another maker’s margin. On the August 4 call the controller said the orbital fleet will run on NVIDIA silicon, “a very significant percentage of their GPUs next year” (REPORTED). So the company will buy the very thing the fab was to make. There is a reading that fits the schedule rather than fighting it: on the county profile the plant spends almost nothing before 2028, so buying chips through the interval is the bridge that timeline implied, not a reversal of it. The cost lives inside that reading, not against it. The essential project’s economics now carry years of buying, at a supplier’s margin in a shortage, the very thing it was meant to make; the fab meant to end the chip bill sits behind a growing one, and the interval prints in the accounts quarter by quarter. That is the cost of not beginning, in a line item. One quarter on, the capstone of the offering has no plan, no money, and no motion. Where did it go?

A hundred billion, raised and not earned

The next need is the cash itself, because a hundred billion dollars had to be raised to face the rest, and it was. SpaceX ended June with about $100 billion of cash and marketable securities, and its chief financial officer, Bret Johnsen, led the August 4 call with the number. By source, it is raised, not earned: about $85.7 billion of offering proceeds and $25 billion of new bonds, against $3.5 billion of operating cash for the half (FILED).

A quarter of it cannot move. The company holds a minimum cash balance of $25 billion, the commitment on which the agencies rest the investment-grade rating. In the public filings the commitment appears only as ordinary language: the company says it aims “to maintain an investment grade credit rating” and “to maintain strong liquidity,” and the reports keep the same posture, but nowhere in them is the number. The size, $25 billion, is Fitch’s, stated in the June 18 rating action (REPORTED); Fitch’s account places the specific commitment in the private bond documentation and the representations to the agencies. So the public record carries the commitment as routine, and a rating agency carried the amount. It is not ordinary by size. Twenty-five billion dollars is a quarter of the cash, about four months of the current burn, and roughly a tenth of the low-growth need we have counted.

This is a shape we have seen before. The roughly $20 billion a year that Google pays Apple to be the default search engine reached the public because an antitrust court compelled it; absent the case it would have stayed dark. The $25 billion here reached the public because Fitch stated it, going further than a rating action had to; absent that, it would have stayed dark too. Two commitments in the same band, each material by a plain reading, each with its size surfaced from outside the filings rather than within them, and each carried as ordinary. That last part is the hard one.

The freeze is the cash cost of the raise. With it off, the deployable figure is nearer $75 billion. At the pace just guided, $18.4 billion of capital a quarter with the next two “very similar,” that is about a year. The fortress raised to close a $235 billion gap is a one-year fuse on borrowed money, a quarter of it held to a commitment the filings carry only as ordinary language, its size known to the public only through a rating agency.

The coupon has begun to print, as we said it would. Interest expense ran $629 million in the quarter, beside the $18.4 billion capital line (FILED), and the $25 billion of notes carry about $1.46 billion a year going forward, the first payment due January 15, 2027, due whether or not the build begins. The notes retired the roughly $20 billion March bridge loan, a debt-for-debt exchange that added no cash the business can deploy, and the half’s accounts carry a $1.2 billion premium for extinguishing the old debt early (FILED). The raise has a mirror in the freeze: the company raised $25 billion in bonds and holds $25 billion in minimum cash, so the sum brought in is the sum that must sit still. A rating was obtained, and its price is a coupon, a premium, and a frozen quarter of the cash.

The use that grows as the price falls

The next is Cursor, agreed at $60 billion, and it is the only item that grows as the stock falls. It is structured as an option: $60 billion paid in stock if it closes, or $10 billion in cash if it does not (FILED). On the live branch, the stock, the price is a fixed value struck in shares at the seven-day average just before closing. Fixed value, floating shares: each down day, more shares are owed to carry the same $60 billion, and the company absorbs all of it. On the August 4 call the controller said the close is coming but that the company is “wary of sort of jumping the gun on regulatory closures” (REPORTED), which holds the window open longer. So for now it is dilution with no cash and no revenue, deepening with each decline and with each week the regulators take, and the decline that deepens it has run alongside the spending it adds. The other branch does not spare the balance sheet. It puts $10 billion of cash back into the gap the moment the stock deal falls through. Neither branch is free.

The only rung being funded

The money reaches the fourth need, and by the rate cash leaves it is the largest of all. Capital expenditure ran $18.4 billion in the quarter, $15.8 billion of it AI infrastructure (FILED), guided flat for the rest of the year on the August 4 call (REPORTED). At that pace the AI buildout alone runs on the order of $300 billion through 2030, more than the whole of the original gap, and it is the reason the deployable cash is a one-year fuse. This is the terrestrial data-center build behind the cloud business, and it pays while it lasts. On the plainest arithmetic the Colossus center rented to Anthropic cost about $13 billion and earns about $15 billion a year, close to a 50 percent return on the build after power and depreciation (REPORTED). That is the bull case, and its flaw is the calendar: a 50 percent return on a contract the customer can end in ninety days while it builds its own capacity elsewhere to replace it.

The company gave this build its own scoreboard. The quarterly introduced a new measure, nameplate compute draw, the installed power the AI fleet can pull, rising from 0.4 gigawatts a year ago to 1.4 gigawatts at June 30 (FILED). The filing states that the figure “reflects installed capacity and does not represent actual power consumption or utilization.” The number handed to the market to track the AI story measures what has been built, not what is used; it rises with capital spent, not with revenue earned or capacity filled. The revenue guide is drawn the same way, forward. On the August 4 call Johnsen put the company on a path to $100 billion of annualized run-rate by December, a single month’s expected revenue carried out to a year, and the controller called it “not a question mark,” reached “if we basically did nothing” (REPORTED). Against $7.8 billion recognized this quarter, about $2.6 billion a month, a $100 billion December run-rate is about $8.3 billion a month, more than triple the current pace in six months. “Nothing” carries the entire ramp landing on time and the ninety-day contracts staying put. It is a run-rate, not revenue booked, and annualized, not recurring. And it answers whether the buildout can be throttled: the controller put $1 trillion of revenue by 2030 on the record, a non-zero chance in 2029, and no path runs from about $31 billion of annualized revenue today to a trillion without the machines that spend buys. The buildout is not the company’s option to defer. It is the revenue promise read from the cost side.

The promise runs both ways, and this is the bull case in full. At $1 trillion of revenue a business of this kind should at last turn a profit and throw off cash, which would close the gap by making the company self-funding. That cash is a potential source, but in the out years and unknown in size now: the $400 billion still has to be spent to reach the scale that produces it, and the spending comes first. Cursor, itself a cash-burning business, adds to the loss before it adds to anything else, and any operating cash the forecast shows is contingent on the buildout landing and the growth arriving. Margins turning sooner would offset the need sooner, though on the filed trajectory that is a late-period event, toward the decade’s end, not a near-term one. One bank’s published estimate has the company burning about $106 billion cumulatively before free cash flow turns positive around 2030 (REPORTED), the same point in numbers: the relief is real, it is late, and it comes only if the ramp lands.

There is a discipline the bull case cannot skip, and it answers the case the company is pressing, that the cash flow to come will be large enough to justify the spend. From an analyst’s chair, revenue and capital move together: this revenue is made of machines, the machines are the $400 billion, and every dollar added to the top line attaches the capital that produces it. A model cannot lift the trillion without lifting the build, and a discounted-cash-flow valuation cannot net a number this large away by discounting the cash it might one day return. The $400 billion is not a cost the future cash flow erases. It is the price of that cash flow, paid first and in full. The whole question is whether the company reaches that scale before the cash and the patience run out, and nothing filed this quarter settles it.

Sixty days, and the plan had moved

There is a fair way to hold all of this, and it belongs in the reading. What the quarter shows is a business plan changing shape sixty days after it was sold. The document that sold the shares in June led with a semiconductor fab and a road to Mars; the first quarter’s cash led with an AI compute build and a new scoreboard to track it, and on the August 4 call the president, Gwynne Shotwell, added a line the offering had not featured, a terrestrial mobile network on the EchoStar spectrum, meant to take “quite a few” customers from the incumbent carriers, whose shares fell that afternoon, and she declined to size its cost (REPORTED). The prospectus permits exactly this. Its forward-looking language reserves the company’s right to change strategy, move capital, and set aside announced projects, and the change may be for the better: if the case for AI is right, pouring the cash into compute and holding the fab back is the correct call, and a market that priced the mission may be slow to reprice the turn. That case could come true. The difficulty is not that the change is wrong. It is that the change came fast, and following it is hard, because it leaves four documents that no longer point the same way. The aspiration is Mars, which gets nothing. The offering is Terafab, which gets nothing. The cash is the AI build, which gets almost everything. The headlines are two different hundred-billions, one of cash raised and one of run-rate projected for a single December, plus $14.1 billion of contracted sales and 1.4 gigawatts of nameplate. Four answers to the one question of what the company is. Sixty days in, a buyer has to choose which to believe.

What is measured, and what is not

That is the question the quarter forces, and we answer it the one way that keeps the reading straight: we do not rebase. The S-1 set out a large aspiration, priced it, and sold it, then laid out, in the ordinary language of what a business does from quarter to quarter, the steps meant to reach it, fixing the expectations, the measures, and the price along the way. The $235 billion gap was measured against those stated steps, and it stays measured against them, because a company changing its mind is the event a fixed baseline exists to catch. The 10-Q was the first ordinary-course report on that plan, and little in it was ordinary: the plan reshaped inside sixty days, eighteen billion dollars of capital in a single quarter, the essential fab gone quiet, a new measure minted to keep score of the turn. The company offers the quarter as routine. What it filed does not read as routine, and a fixed baseline is how a quarter like that is seen for what it is. So the gap holds, and it grows. What we track from here is the distance between the two sets: the things the offering promised, and the things the cash chose. The company may change what it builds. It does not get to change what it is measured against.

That is what the price is paying for. The premium above the standing businesses, the part no model reaches, is the value of the attempt, the option that one of these bets comes good and the company gets to do the thing it was sold to do. Trying is what holds the valuation up, and the cost of not beginning is that the option decays as the attempt recedes, quarter by quarter, pulled forward on the slide and pushed back in the cash.

The Moon is about 239,000 miles away, the mid-term step toward the mission. At the original gap that was about a million dollars a mile; at today’s it is closer to a million seven, and the quarter spent eighteen billion and closed not one of them. Anyone who bought this to fund the mission bought a plan for a plan for a plan, each of which must go right before the company earns the right to attempt the first. The gap did not narrow this quarter. It widened, from about $235 billion to about $400 billion, the money went to the fourth thing on the list, and the first thing, the essential one, went quiet.

Related, on the Quality of Cash shelf: The $235 Billion Cash Gap; The Ninety-Day Annuity; The Cost of Not Beginning; Cursor Stock and the Investment-Grade Refinancing; The Terafab Record; The Probability of Failure; The Three Layer Cake: SpaceX’s Governance Structure.

Standing disclosure

Cape Fear Advisors holds no direct position, long or short, in the securities discussed here. Any exposure is indirect, through managed funds it does not control, which may now include index funds holding SpaceX. Anthropic is the developer of Claude, which is used in preparing this research, and Anthropic is also a compute counterparty to SpaceX, the customer whose contract is read here. That nearness cannot be fully checked away, which is why no claim here rests on trust in the tool: every figure carries a public source, and the record grades the rest. Others named have ties to Anthropic, among them Google, a second compute counterparty, and its parent Alphabet, an Anthropic holder; and companies not named here may hold positions or supply relationships that bear on the filers discussed, which is why every piece is re-checked for bias, ground facts, and filings rather than read against a fixed list. Figures are quoted from the filers and from named parties without characterization, and the same standard of reading is applied to every party named.

NOTES

The gap, and the exhibits. The commitment stack and the sources against it are our own common-currency construction, not a filed figure; the components are filed, and each is tagged below. The total of about $235 billion is our estimate from “The $235 Billion Cash Gap“ (May 21, 2026), which enumerated it without a single table; the May column here restates that enumeration on the common cash-to-be-deployed-through-2030 basis, with spectrum shown at its cash component and the bridge moved to the sources side. That piece modeled a raise of $50 to $75 billion gross; with the over-allotment the offering reached about $86 billion gross. The updated total of about $400 billion holds the same items on the same basis and carries the AI line at the guided pace; the increase is the AI line. The May stack carried Terafab at $55 billion, the Phase 1 figure; the update carries about $5 billion, the county schedule through 2030, and the AI line’s increase is larger than the total’s because the Terafab line fell. The $400 billion is a gross figure, the cash to be deployed. Operating cash flow is a potential offset in the out years, unknown in size and arriving after the spend; it does not lower the amount that must be deployed, only, on a lag, the outside cash needed to reach self-funding (see the crossover estimate below).

The quarter. SpaceX Form 10-Q for the quarter ended June 30, 2026 (accession 0001628280-26-052535): quarter revenue $7.81 billion (half $12.5 billion); capital expenditure $18.4 billion for the quarter, of which $15.8 billion AI-segment (half $28.5 billion); interest expense $629 million in the quarter (half $1.29 billion); operating cash flow $3.5 billion for the half; cash and marketable securities about $100 billion at June 30 ($93.5 billion cash, $6.5 billion marketable); nameplate compute draw 1.4 gigawatts at June 30 against 0.4 gigawatts a year earlier, with the filing’s own note that it “reflects installed capacity and does not represent actual power consumption or utilization.” FILED. The $14.1 billion of contracted sales is the company’s cloud-agreements bookings highlight from the Q2 release, and is not GAAP backlog or recognized revenue. REPORTED. The $100 billion annualized run-rate target for December and the $1 trillion revenue target for 2030 (with a non-zero chance in 2029), the “not a question mark” and “if we basically did nothing” phrasing, the “a very significant percentage of their GPUs next year,” the “wary of sort of jumping the gun on regulatory closures,” and the president’s terrestrial-network remarks are from the Q2 2026 earnings call, August 4, 2026. REPORTED.

Terafab. The “that’s why we need to do Terafab. It seems essential” quotation is from the June 4, 2026 video interview of Elon Musk by Jamie Dimon, filed as a Rule 433 free writing prospectus (accession 0001628280-26-041365, transcript at Exhibit B), and the roadshow free writing prospectus (accession 0001628280-26-040610). FILED. The up-to-$119 billion build scale and the $5 billion-by-2030 development-agreement figure are from the Grimes County, Texas JETI tax-abatement filing (May 6, 2026) and the associated development agreement. REPORTED, county records. The by-year spend profile (near zero before 2028) is our consolidation of the county schedule, read in “The Terafab Record.”

The cash, the freeze, and the coupon. The IPO closing Form 8-K (accession 0001628280-26-043288) records 638,888,888 Class A shares sold at $135.00, about $86.25 billion gross and $85.7 billion net after fees, with the over-allotment exercised in full. The $25 billion senior notes are in the notes Form 8-K (accession 0001628280-26-044955), five tranches, weighted average coupon about 5.855 percent, about $1.46 billion of annual interest, first payment January 15, 2027. The 10-Q for the half records proceeds from debt and other financing of $51.8 billion, repayments of $39.4 billion, and a $1.2 billion debt-extinguishment premium; the notes retired the roughly $20 billion March 2026 bridge loan, a debt-for-debt exchange that added no deployable cash. FILED. The $25 billion minimum cash balance: the commitment appears in SpaceX’s public filings as general language, the aim “to maintain an investment grade credit rating” and “to maintain strong liquidity” (S-1 and 10-Q, FILED); the specific figure, $25 billion, is stated in the Fitch rating action of June 18, 2026 (REPORTED), whose account places the commitment in the private bond documentation and the representations to the agencies. The finding here is that provenance: the filings carry the commitment as general language, and the size is Fitch’s. Earlier entries in this series carried the $25 billion as committed in the filings; with the first public 10-Q now in hand, we refine that language against the new filing, and the commitment’s size and effect are unchanged. The roughly $20 billion Google-Apple default-search payment cited for comparison is from the United States antitrust litigation. REPORTED.

The crossover. The estimate of about $106 billion of cumulative cash burn before free cash flow turns positive around 2030 is HSBC’s, reported through coverage (July 24, 2026), and read in “The Cost of Not Beginning.” REPORTED.

Cursor. Disclosed before the offering and read in “Cursor Stock and the Investment-Grade Refinancing“; the merger is recorded in Form 8-K (accession 0001628280-26-043411), with the $60 billion stock exercise or $10 billion cash alternative and the seven-day pricing average carried in the 10-Q (Note 20). FILED.

Valor and spectrum. Both appear on the exhibit and are carried here for completeness. The Valor equipment leases, about $20 billion recorded as failed sale-leasebacks with a related party, are in the 10-Q (Note 17). The spectrum purchase from EchoStar is about $11.5 billion of cash, of a $19.6 billion total, in the 10-Q (Note 6); the $11.5 billion is the spectrum purchase price, and the cost of the terrestrial network the president described on the August 4 call is separate and unstated. FILED for the purchase and lease figures; REPORTED for the network remarks.

Colossus economics. The approximate $13 billion build cost and about $15 billion of annual revenue on the Anthropic rental, and the resulting return, are reported estimates read against the filed contract terms, not a filed figure. REPORTED.

The registration statement. SpaceX Form S-1 (accession 0001628280-26-036936) and final prospectus 424B4 (accession 0001628280-26-042639), for the forward-looking language permitting a change of plan and the risk-factor language on future issuance (”further issuances of equity or convertible debt securities” and “significant dilution”). FILED.

Analysis: Cape Fear Advisors.

The question, and the tense it comes in

The question this week, asked everywhere either stock is discussed, is whether Tesla and SpaceX will merge, and when. Held against the filed record, the question is mis-timed. The merger is already happening, in most of the ways two public filers can combine, and each of those ways is documented below. One word gets its two meanings separated at the door: in the statute books, a merger is a specific act, a vote and a certificate that no one has filed, and nothing below says otherwise; that formal act is called a combination throughout this piece. In the market’s mouth, the merger is a condition. The title borrows the market’s word, and the piece reports the condition’s tense. If the structure of it looks malformed, a project with no name in one filing and no numbers in the other, a stake that arrived sideways, an influence presumed from a person, the form is largely an outcome of the several rulebooks the arrangement balances and flows through: securities disclosure that captures events, accounting that captures elections, related person policies that capture transactions, state law that sets the venue. Each instrument took the shape its rulebook allowed. The shape of the whole is what a combination looks like while it remains unfiled. What remains undone is the formal transaction, and no one outside the companies knows what steps would be taken or how; nothing here predicts any. What can be done from outside is read the state of the merger as it stands: what has crossed, what it did to the financial statements, what the boundary costs to keep, and what a formal combination would change about what a holder of either stock can see.

The one person who can settle the question was asked it directly on July 22, on Tesla’s second quarter earnings call, and answered in two parts. The first described the present: “there’s more and more overlap.” The second deferred the future: combining companies “can’t” be discussed on an earnings call and “has got to be done with the appropriate process.” Transcript providers differ on the exact wording, and the Notes carry the variants. The two parts of the answer are the two halves of this piece: the overlap, present tense and filed, and the process, future tense and not. One more thing the record does not show is any requirement that the state resolve. Entangled and separate is itself a durable configuration; controllers have run constellations of related public companies for decades without combining them, and nothing in either filing sets a clock on the arrangement itself. The present tense of this piece is not a countdown, and the one clock in this story belongs to its owner.

One project, filed in halves

Start with the chip project, where the market’s merger logic starts.

SpaceX’s prospectus names the project and declines the numbers. Terafab appears in the S-1 as “a chip manufacturing initiative with Tesla and Intel,” governed by “a general framework” whose financial terms, intellectual property rights, and ultimate term are, in the filing’s own words, not finalized. A sentence the Commission staff’s comments produced, repeated at four separate places, states that any specific projects under the framework “have not yet been determined.” The drafting history is part of the record: the confidential draft of March 30 described “an announced chip manufacturing initiative in partnership with Tesla,” with no framework caveat and no mention of Intel; the staff’s first comment letter forced Intel in by name and asked for material terms, timelines, milestones, and capital expenditures; the second pressed the company to state plainly that none had been determined. What survived into the final prospectus is a project with named partners and no determined content. Both versions are FILED, with accessions in the Notes.

Tesla’s most recent quarterly filing does the opposite: the activity and never the name. The management discussion states that Tesla is “expanding our scope of manufacturing to include semiconductor and solar fabrication.” The word Terafab appears nowhere in the document and has never appeared in any Tesla filing; the count since March 2026 is zero, through a full quarterly report and the earnings release beside it. The zero broke only in an unfiled venue: on the July 22 call the controller discussed Terafab by name, called it necessary, and said a location announcement was expected soon.

Between the two filings, the project is fully described by neither. The specific dollars live in a third place entirely, a Texas county and state record, under the name of a SpaceX subsidiary, describing phased investment of up to $119 billion, read in full earlier in this series in The Terafab Record. One company filed the name. The other filed the activity. The dollars sit in a county clerk’s office. And the Texas record carries the merger thesis in miniature: the May applications describe the project, in a sworn filing, as “led by a consortium of affiliated advanced technology companies, including Tesla, Inc., Space Exploration Technologies Corp. (SpaceX), and xAI Corp.” By the swearing, xAI had not existed as a separate company for three months. SpaceX’s own prospectus combines it retrospectively, effective February 2, 2026, and never uses the name xAI Corp at all. The sworn description names a partner the paperwork had already absorbed; the name outlived the company, which is this piece’s argument about the boundary, written into the record by the project itself. A reader who wants the whole project must consolidate three records that no single filer presents together.

The staffing runs ahead of the paperwork. In June, a seventeen year veteran of Intel’s fabs, most recently the factory manager responsible for Intel 18A technology development transfer, construction, tool installation, and startup, joined Tesla in Austin. His title, per his own verified profile, is Director, Terafab. None of the three companies announced the hire, and no filing carries it; it became public because reporters read his resume. The word that appears in no Tesla filing is a job title on Tesla’s payroll, held by an executive supplied, in effect, by the third partner, for a project the second partner’s prospectus says has not been determined. REPORTED; sourcing in the Notes.

The counterparties themselves answer differently on whether the project’s name exists. Across the July prints: ASML named Terafab on July 15; TSMC did not on July 16; Tesla filed zero and said the name on its call on July 22; Intel, whose chief executive named the partnership in April, said the word zero times on its own July 23 call, while presenting the roadmap for exactly the node class the project’s sworn application names. One project, four public companies touching it, four different answers to whether it can be said out loud. Intel has itself filed nothing on the project, ever.

And the dependency runs backward across the wall. On Tesla’s own call, the controller said the chip project is necessary to scaling Optimus, which is Tesla’s product roadmap. That statement is oral and REPORTED; it appears in no Tesla filing. Tesla’s next act, by its controller’s own words in Tesla’s own venue, depends on a project whose name Tesla has never filed and whose determined content does not exist in the partner’s prospectus.

The money already crossing

The equity crossed in March. Tesla’s related party note: “Upon receiving the applicable regulatory approvals, the Company invested $2.00 billion in SpaceX common stock (formerly a preferred share investment in xAI) representing an ownership interest of less than 1% in March 2026.” Read slowly, that sentence documents a chain: Tesla invested in xAI; SpaceX acquired xAI and paid in SpaceX’s own stock; Tesla’s position converted. The stake arrived as change from a transaction between two other rooms owned by the same person. The earlier entry in this series observed that SpaceX’s acquisitions have been paid for in its own stock, xAI and Cursor both; Tesla’s footnote is the counterparty record of that currency, received and now marked, by a filer sharing the same controller. FILED.

The commerce crossed all year, and it grew. The same note discloses that SpaceX bought $318 million of Tesla’s Megapack products in the second quarter, against $242 million of cost, in the ordinary course, under Tesla’s Related Person Transactions Policy. The six month figures put the first quarter near $87 million, so the second quarter ran about three and a half times the first, and the single related buyer supplied about one tenth of Tesla’s energy generation and storage revenue. The prior year’s comparative is the sentence after: transactions with related parties were immaterial in 2025. This commerce did not exist at scale five quarters ago. Grid storage is terrestrial data center infrastructure, and SpaceX’s own prospectus attributes its capital spending surge to “the rapid expansion of our terrestrial data centers,” so the buyer’s capital line and the seller’s revenue line are watching the same buildout from opposite sides of one invoice. Both materiality determinations are the issuers’ own and defensible at these sizes; the observation is only that the money is crossing, at a growing rate, and that the record says so.

And it crosses in one direction. Nothing in the record read here shows SpaceX holding any Tesla stock: a position above five percent would file its own disclosure and none exists, and the offering documents’ balance sheet, read in full earlier in this series, carries nothing below it. The paper flows one way, outward from the company whose stock is the currency, and the sensitivity flows with it. Tesla’s income statement is now a function of SpaceX’s tape. SpaceX’s owes nothing to Tesla’s.

Three incomes, and the one that carried the quarter

Then the mark, which is where the entanglement stopped being a footnote and reached an income statement the whole market reads. Olga Usvyatsky surfaced the Tesla side of this on the day the records diverged, in her July 26 analysis of how Alphabet and Tesla accounted for the same investment differently, and the accounting argument that follows is hers, cited with her name on it.

Tesla’s policy note states that Tesla is “presumed to have significant influence over our equity method investment in SpaceX under ASC 323... as our CEO also serves as the CEO of SpaceX,” and that, not having control, Tesla “elected the fair value option in accordance with ASC 825... to provide a more relevant measure of the investment’s current economic value.” The consequence, as Usvyatsky wrote and as Francine McKenna quoted her: under the equity method Tesla would have recorded its proportionate share of SpaceX’s operating results, and SpaceX’s most recent filed quarter is a net loss of $4.28 billion. The fair value option instead remeasures the stake and runs the change through income. The economics did not move an inch; the accounting produced about a billion dollars of pre-tax income. At an ownership interest under one percent, the share of losses avoided would have been tens of millions; the swing is almost entirely the gain recognized. Both treatments are permitted. Tesla chose, and said why in one filed sentence.

What the choice did to the quarter reads best as three lines. In the three months ended June 30, Tesla’s operating income was $398 million. Its interest income was $422 million. Its unrealized gain on the SpaceX stake, essentially all of which belongs to the quarter on the bridge set out in the Notes, was $1,005 million, against pre-tax income of $1,329 million. Three incomes: the register, the vault, and the bet. The vault out-earned the register. The bet out-earned both combined, and supplied about three quarters of the quarter’s pre-tax income. Revenue grew 25 percent from the prior year while operating income fell nearly 57 percent, so the six month bottom line, flat against last year at $1.6 billion, is flat only because a mark on the neighbor replaced what the operating business used to earn; about half the six months’ pre-tax income is the one line.

The election has a sequel every quarter, and the sequel is running now. Tesla’s measure is taken from market observable inputs, and SpaceX’s stock has fallen hard since the June 30 measurement date; Usvyatsky flagged earnings sensitivity to the SpaceX share price as the coming issue before this week made it visible. On the same convention, scaled to the quote, the position that measured near $3.0 billion at June 30 stood near $2.0 billion at Friday’s close, near its cost: the billion dollar gain that carried the second quarter existed for about four weeks of tape. The exact figure belongs to Tesla’s next filing; the direction and the rough size belong to arithmetic. This week’s coverage explains one stock by the other, a drag, a tumble, a pull. The filed mechanism is plainer than the metaphors: an election converts one company’s tape into the other company’s income statement, in both directions, mechanically, every quarter. Nothing is dragging anything. A footnote is doing exactly what it said it would do.

The pile, and its price

Beside all of this sits the cash. Tesla held $15.2 billion of cash and equivalents and $28.3 billion of short term investments at June 30, about $43.5 billion together, roughly flat since December. Across the six months the company generated $8.6 billion of operating cash flow and spent $8.3 billion on capital expenditure, so the pile neither grew nor deployed; it earned $422 million of interest in the quarter, an annualized yield near 3.9 percent.

One asset sits beside the pile and belongs to none of it. The SpaceX stake is presented within other non-current assets, not among the cash or the investments, so nothing above includes it. It is an asset that produces income and no capital: it ran a billion dollars through the income statement while sitting unspendable, since the shares predate the offering and are of the class the lockup agreements cover, the measurement runs on market observable inputs rather than the exchange quote, and a sale of size by this particular holder would be a related party adjacent event in both companies’ records. Income statement active, balance sheet inert. Held against the system the two companies already form, the stake nets to nothing even now: one room’s asset is the other room’s paper, and a system cannot spend a claim on itself. The gain that carried Tesla’s quarter was real to Tesla’s holders. To the system, it was the system marking itself.

Across the wall, the room with the attempt priced $25 billion of notes in June at a weighted coupon of about 5.86 percent, and committed in its filings to hold $25 billion of minimum cash that the ratings rest on. The controller’s system, read across the two filings, is paid about 3.9 percent for money in one room and pays about 5.9 percent for money in the other, and the wall between the rooms is the corporate separateness whose door the July 22 answer named. That is an observation about two filed records side by side, and it is left as one.

Tesla is anything but inert; the opposite problem marked the quarter, with capital spending surging on robotics, on artificial intelligence infrastructure, and on a development fab in Austin that is Tesla’s facility and Tesla’s spending. The precise statement is narrower. The one attempt Tesla’s roadmap depends on, by the controller’s own words in Tesla’s own venue, is the one thing the $43.5 billion is not funding, and the one word Tesla’s filings never use. The company is trying at every scale except the one the dependency requires. Its cash earns a money market yield while the project its next product requires is financed next door at a premium, by a company that had to borrow. None of that is an allocation critique: Tesla capital into the fab would itself be a related party transaction of scale, so the kept boundary and the undeployed pile are the same fact seen twice.

The boundary, and its keepers

All of this crossing is managed, and the management is visible, which is itself information. Tesla’s related party note invokes the Related Person Transactions Policy by name. On the July 22 call, within moments of the controller’s merger answer, Tesla’s general counsel described SpaceX as a “great partner” providing “numerous beneficial transactions,” the boundary being tended in real time, in the venue where the question landed. The significant influence sentence is careful drafting: influence presumed, control expressly disclaimed, the election justified in a single clause. On the other side of the wall, SpaceX’s prospectus discloses the controlled company structure in full: two classes, one vote and ten, about 85 percent of the voting power on about 42 percent of the economics, a board exempt from independence requirements, a shareholder proposal threshold the vote structure places out of reach. None of the apparatus is improper. Its visible effort is the point: this much machinery exists because this much boundary needs keeping, and the keeping is a cost both shareholder bases pay every quarter.

There is a filed history of what “the appropriate process” means at this address. In 2016 the controller stood on both sides of a combination once before, when Tesla acquired SolarCity. That transaction produced a process, the process produced six years of litigation under the entire fairness standard, a trial court ruling in 2022, and an affirmance in 2023. The history predicts nothing here. It is cited because it is the one documented instance of the machinery the July 22 answer invoked, and because it prices, in years and in court records, what controller on both sides costs to clean.

That history has a jurisdiction problem, and the jurisdiction may be the least examined player in the whole question. The SolarCity process ran in Delaware, under Delaware’s entire fairness standard, in a court with a century of precedent on controllers. Neither company is a Delaware company anymore. The sequence: in January 2024 the Delaware Court of Chancery voided the controller’s Tesla compensation award. Within weeks, SpaceX reincorporated in Texas. By June, Tesla’s shareholders had approved the same move. In September 2024 the Texas Business Court opened, a new venue created by the 2023 legislature for large commercial and governance disputes. Tesla’s Texas governing documents carry a 3 percent ownership threshold for derivative claims, which at current prices asks a would-be plaintiff to hold tens of billions of dollars of stock before bringing the kind of claim the SolarCity holders brought; in 2025 the state codified presumptions favoring directors and officers and authorized thresholds of that size. SpaceX’s package is reported to run further, to mandatory arbitration of shareholder claims. Even the listing participates: the offering was admitted to Nasdaq and to Nasdaq Texas, per the prospectus cover. Each step was lawful and disclosed, and the sequence is stated as sequence. What it produces is a plain fact about the last rung of any ladder: the one precedent for the appropriate process was made under a standard, in a courtroom, that no longer governs either company, and the courtroom that does govern them is two years old, thinly precedented, and has never priced a transaction like the one the market keeps asking about. The SolarCity record prices the old venue. The new venue has no prices yet.

And beneath the apparatus sits the fact the apparatus cannot reach. The significant influence presumption in the accounting literature ordinarily arrives through ownership: a fifth of the shares, seats on a board. Tesla holds less than one percent of SpaceX and presumes significant influence anyway, for the reason its own footnote gives: the CEO is the CEO. The stake does not create the influence; the person does, and the filing says so. Policies can wall transactions, committees can wall approvals, disclosure can record what crosses. Nothing partitions the place where the decisions of both companies form. The accounting has already located the merger where it lives.

The ladder, and its paperwork

What would deeper entanglement look like, and how would a reader know? Every rung has a filing attached, so the forward looking content of this piece can be a list of documents rather than a forecast.

A definitive Terafab project agreement, the step that converts the framework’s undetermined projects into determined ones, files as a material definitive agreement at SpaceX and, at the sizes the Texas record describes, at Tesla; its absence to date is why the prospectus sentence still governs. Tesla capital committed to the fab, in any form, files as a related party transaction, with the proxy scale disclosure that follows size. A combination itself requires the process the controller named, a registration statement, and the fairness machinery whose one local precedent ran six years.

The last rung carries one more set of documents, and this series read them in May, before the offering existed. Both companies adopted 2025 performance awards, and both plans carry change in control provisions. Under Tesla’s plan, a qualifying transaction disregards the operational milestones, the million robots and the million robotaxis and the $400 billion of adjusted EBITDA, and the deal price itself satisfies the market capitalization tranches of an award of up to 423.7 million shares. SpaceX’s award sits in 200 million super voting Class B shares, in a company where the Class B holders alone can remove the chief executive. The mechanics were set out in The $8.5 Trillion Grease Gun on May 1, which also noted that the architecture, once the offering completed, could not be undone; the offering completed on June 12. No scenario from that piece is repriced here and no intention is attributed to anyone. The record holds two filed logics attached to the same act: an industrial one, the fab both roadmaps need, and a compensatory one, a transaction that converts two award structures with the operating tests waived. Both are in documents. They are stated side by side and left there.

One more thing the ladder shows, against the market’s intuition: the rungs are not equally heavy, and the heaviest is the one treated as the modest option. A definitive joint venture, the step the framework anticipates, is the most governance-expensive structure two companies sharing a CEO can operate. Every project the framework ever determines becomes a negotiated contract between related parties, and the prospectus’s own list of what stands unresolved, the financial terms, the intellectual property rights, the ultimate term, is precisely the list of conflicts needing management, project by project, for as long as the venture runs. Each determination would ask for independent process on both sides of the table, and the one person who holds the whole plan in his head could represent neither side of it. The framework asks the wall to hold at exactly the place this piece has already shown no wall can be built. A combination would not meet that requirement. It would dissolve it. Dissolving it is not lightness: a combination concentrates the same complexity into a single transaction, with the process, a new venue, and two award structures of its own to carry. None of the rungs is light. The ladder sorts the shape of the weight, recurring against concentrated, not its presence. Whether any of that bears on which rung ever gets climbed is not knowable from outside and is not guessed at here. What is knowable: at least four months after the framework was announced, the number of projects determined under it is zero, and the structural weight of determining any is part of the record a reader can price.

The ladder also has a clock, and the clock belongs to no one writing about it. The owner spoke it on July 22: a location, soon. His prospectus set no date for the fab, and needs its chips for a constellation it expects as early as 2028. His county agreement pays $10 million on August 2 and dates its investment commitments to 2030, and the product he says needs the chips is Tesla’s next act. Around the undetermined framework the parties are not evenly placed: one company has the money, the engineering bench, and the demand; one has the county agreements and the payment falling due; the named third partner has the node and has filed nothing; and the machine suppliers are further along in speech than the partners are in paper, with ASML naming the project on its own call while the framework that would order its machines remains undetermined. A fab, unlike a holding structure, does not keep. So the durable constellation reading meets its limit here: the configuration can persist, but the project cannot proceed inside it as it stands. The framework must resolve into some form, determined projects, a different structure, or one partner proceeding alone, and each form leaves a different mark in the record. If Terafab moves forward inside the framework, something formalizes and files. If it moves forward outside the framework, one partner proceeding alone, the Commission’s files stay silent, exactly as the previous entry read that silence, and the mark lands in the Texas record instead: permits, payments, construction, the clerk’s side of the ledger. If nothing moves, the dates pass and say so. Three roads, three different papers, all on the owner’s clock and no one else’s.

What a combination would make invisible

The market’s question carries an assumption: that a merger would clarify things. The record runs the other way, and the other column comes first, because the dark would not be bought for nothing.

The savings would be real. One board where two now each run their own process. One disclosure apparatus where two now tend a boundary every quarter. No related person machinery standing between a battery and its buyer, no arm’s length price constructed for a transaction whose two sides already share a mind, no general counsel following the controller’s answers in real time, and no venture whose every project requires the one person who understands it to recuse from both sides of it. The entanglement documented above carries a standing maintenance tax on both companies, and a combination cancels the tax. No figure exists outside the companies, but the direction is not in doubt: the case for combining is a real case, and it is a case for efficiency.

What it costs is the record itself.

The marks would go away. A parent cannot hold a mark to market gain on its own subsidiary. The $1,005 million, the line that made the second quarter and is unmaking the third, ceases to exist by operation of consolidation, and with it goes the most legible signal the entanglement has ever produced. The mark made the relationship visible on a page the whole market reads, and the mark is precisely what a combination deletes.

The cash would go invisible. The batteries would still ship; the flow would not stop. But an intercompany sale nets to zero in consolidation, and the disclosure that today lets anyone compute the margin, $318 million against $242 million, about 24 points, dies with the eliminations. Whatever shadow survived would survive in segment reporting, and segment reporting runs on the management approach: the segments are what the chief operating decision maker reviews, and intersegment revenue appears only if it sits inside the measure that officer looks at. The chief operating decision maker of the combined company would be the controller. Whether a reader ever again saw the flow between the rooms would be, in the most literal sense the accounting framework allows, his judgment call, made by choosing what he reviews. That is a description of the framework, not a legal conclusion, and the Notes say so. This series has spent a month on materiality by silence, the judgment a company renders by filing nothing. Consolidation would hand the record its successor: visibility by judgment.

The stake would end the strangest way of all. Its final mark would be set by the transaction itself, the deal price being the last quote it is ever measured against, and then it would stop being an asset entirely, because a combined company holding the stake would be holding its own shares. An asset that was never available to the combination nets to nothing inside it. The mark is priced by the deal and extinguished by the closing, in the same instant.

The acknowledgments would dissolve. The significant influence sentence becomes control and disappears into a parent’s consolidated whole. The related party note, the policy citation, the general counsel’s tended boundary, all of it becomes internal, in the way that a wall becomes a hallway.

So the state of the merger is this. It exists in a framework with named partners and no determined projects. It exists in a job title on one company’s payroll spelling a word that company has never filed. It exists in a stake that arrived as change, in an invoice that grew three and a half times in a quarter, in a mark that carried one quarter and is unmaking the next, in an influence presumed from a person rather than a position, and in two award structures waiting at the last rung with the tests waived. It exists, most completely, in the only consolidated entity the record currently has: a reader holding both filings at once. What it does not yet exist in is a filing of its own, and the record makes one thing plain about that final document: its arrival would not turn the lights on. The lights are on now. The filings are the lights.

A holder of either stock is currently paying for the boundary and collecting the visibility. The wall costs process and buys sight. The dark costs sight and buys speed.

One filed sentence sits underneath that trade, and the previous entry in this series rested on it. SpaceX’s prospectus states that “we believe speed is a competitive advantage,” and that the company prioritizes “execution speed, capacity expansion, and technological leadership over near-term margin optimization.” Yesterday’s piece read that doctrine as the plan’s engine, and the governance structure as its stated price. Against the trade above, the sentence does one more job: the dark buys speed, and speed is the one good in this trade that one of the two companies has already declared it values above the others, in a registration statement, to the buyers of the offering. The doctrine does not answer the question of the last rung. It prices one side of it, in the issuer’s own words, and it has been on file since May. The doctrine speaks to building rather than to dealmaking, and carrying it from one to the other is this piece’s transfer, not the filing’s claim.

The record supplies the price list for both sides of the trade, and the choice between them belongs to a process one person would begin, in a courtroom with no precedents, on a date no filing yet names.

Related, on the shelf: SpaceX, Adding It Up: The Cost of Not Beginning; The $8.5 Trillion Grease Gun; The $10.25 Trillion Hypothesis; The Terafab Record; The Three Layer Cake: SpaceX’s Governance Structure; The $235 Billion Cash Gap; Cursor Stock and the Investment-Grade Refinancing. On the Tesla mark, read Olga Usvyatsky first; on the income mechanics of unrealized gains, Francine McKenna.

Standing disclosure

Anthropic is the developer of Claude, which is used in preparing this research, and Anthropic is also a compute counterparty to SpaceX, whose contracts are part of the record read here. That nearness cannot be fully checked away, which is why no claim here rests on trust in the tool: every figure carries a public source, and the record grades the rest. Others named have ties to Anthropic, including Google and its parent Alphabet, a SpaceX compute counterparty and a large holder of Anthropic; and companies not named here, among the chip and cloud suppliers, may hold positions or supply relationships that bear on the filers discussed, which is part of why every piece is re-checked for bias, ground facts, and filings rather than read against a fixed list. Figures are quoted from the filers and from named analysts without characterization, and the same standard of reading is applied to every party named.

NOTES

How to read the labels. Every load-bearing figure below carries a source and a status. FILED means a document filed with the Commission and cited by accession, which any reader can pull. REPORTED means the public record outside the Commission’s files: transcripts, local and trade coverage, market data, and analysis posted by its author. The distance between those two words is, as ever, much of what the piece is about.

Prices and figures. Tesla figures are FILED, from the Form 10-Q for the quarter ended June 30, 2026 (accession 0001628280-26-049270) and the Form 10-Q for the quarter ended March 31, 2026 (0001628280-26-026673): revenue $28,236 million against $22,496 million a year earlier; operating income $398 million against $923 million; interest income $422 million; income before income taxes $1,329 million for the quarter and $2,077 million for the six months, with a quarterly provision of $201 million; net income $1,128 million for the quarter and $1,619 million for the six months against $1,610 million; cash and equivalents $15,219 million and short term investments $28,305 million at June 30 against $16,513 million and $27,546 million at December 31; six month operating cash flow $8,634 million and capital expenditure $8,282 million; energy generation and storage revenue $3,139 million for the quarter. Market prices are REPORTED, closes only, intraday prints excluded: Tesla closed $313.03 and SpaceX $115.07 on Friday, July 24, 2026. Percentages are rounded against the argument throughout: the mark is 75.6 percent of quarterly pre-tax income, stated as about three quarters; 48.4 percent of six month pre-tax income, stated as about half; the Megapack revenue is 10.1 percent of segment revenue, stated as about one tenth; interest income annualizes to 3.88 percent on the June 30 pile, stated as about 3.9.

The mark, and the election. FILED, Tesla 10-Q. The policy language verbatim: “We are presumed to have significant influence over our equity method investment in SpaceX under ASC 323, Equity Method Investments and Joint Ventures, as our CEO also serves as the CEO of SpaceX but as we do not have control over the investee, we have elected the fair value option in accordance with ASC 825, Financial Instruments, to provide a more relevant measure of the investment’s current economic value to financial statement users.” The related party note verbatim: “Upon receiving the applicable regulatory approvals, the Company invested $2.00 billion in SpaceX common stock (formerly a preferred share investment in xAI) representing an ownership interest of less than 1% in March 2026.” The cash flow statement carries the $1,005 million unrealized gain for the six months; the first quarter filing shows the position acquired for $2,002 million in the first quarter and carried at approximately that amount at March 31, then classified within Level 2, so the attribution of essentially the whole gain to the second quarter is a bridge across the two filings rather than a printed quarterly line. The reversal arithmetic in the text scales the June 30 measure proportionally to the SpaceX quote and is an estimate on a stated assumption; the filed answer belongs to the third quarter report. The comparison of Tesla’s accounting with Alphabet’s, and the argument that the fair value election rather than any valuation input is the choice that produced the income, belong to Olga Usvyatsky (“Tesla’s investment in SpaceX: The $1 Billion Accounting Choice,” July 26, 2026), quoted publicly by Francine McKenna; her work is the reason this piece exists in its present form. One shape note is this piece’s own arithmetic on her point, not hers: at under one percent ownership, the equity method share of losses avoided is tens of millions, so the billion dollar swing is almost entirely the gain recognized. The S-1 states the same conversion from its side, describing Tesla’s “January 2026 commitment to invest in xAI” as an investment that, upon SpaceX’s acquisition of xAI, “was converted into an equity interest in SpaceX.” Both filers describe the same change of hands, independently, from opposite ends of it.

The framework, and its drafting history. FILED. The confidential draft registration statement of March 30, 2026 (accession 0001628280-26-021860) described “an announced chip manufacturing initiative in partnership with Tesla,” without Intel and without caveats. The staff’s first comment letter (0000000000-26-004247) directed the company to discuss the material terms of Intel’s partnership, timelines, milestones, and capital expenditures; the second (0000000000-26-005119) pressed the company to state that these “have not yet been determined.” The S-1 and final prospectus (0001628280-26-036936; 0001628280-26-042639) carry the surviving language: a chip manufacturing initiative with Tesla and Intel, a general framework, terms not finalized, and, in four places, “Any specific projects undertaken pursuant to this framework will be subject to separate negotiations and agreements (including any development timelines, milestones and capital expenditures) and have not yet been determined.” The Texas record, REPORTED from county, district, and Comptroller documents and local coverage, is read in The Terafab Record and carries the up to $119 billion of phased investment under TeraFab AI, LLC. The applications’ Parent Company Information names Space Exploration Technologies Corp alone, at the same 1 Rocket Road address as the applicant; no xAI entity appears on any ownership line of the eight sets. The sworn project description in the May applications reads “led by a consortium of affiliated advanced technology companies, including Tesla, Inc., Space Exploration Technologies Corp. (SpaceX), and xAI Corp.” The S-1 presents its financials as the retrospective combination of the companies to include the historical results of xAI, “which was acquired by SpaceX, effective February 2, 2026,” the entity history running through X.AI Holdings Corp., which had itself absorbed X Holdings on March 28, 2025; the prospectus uses X.AI Holdings Corp. and later X.AI Holdings LLC, and never the name xAI Corp. The applications’ node claim, sworn: “sub-2 nanometer class nodes.”

The staffing. REPORTED. Gary Jiang’s verified LinkedIn profile, as identified by Electrek on June 30, 2026 and covered by Tom’s Hardware, TrendForce, and others through July 1, lists 17 years and 9 months at Intel, most recently as factory manager responsible for Intel 18A technology development transfer, construction, tool installation, and startup, with earlier fab management at Ocotillo, and a June 2026 start at Tesla in Austin under the title Director, Terafab. Coverage varies between seventeen and eighteen years; the profile’s figure is used. No company announcement or filing exists.

The word counts. Tesla’s FILED Terafab count is zero across all filings since March 2026, including the July 22 earnings release exhibit (8-K, 0001628280-26-049213) and the July 23 10-Q. The 10-Q’s management discussion states Tesla is “expanding our scope of manufacturing to include semiconductor and solar fabrication.” The four print scoreboard is REPORTED from earnings transcripts and coverage: ASML named Terafab on July 15; TSMC did not on July 16; Tesla said it on the July 22 call while its filings stayed at zero; Intel’s chief executive discussed the partnership by name on April 23 and the word did not occur on Intel’s July 23 call, including in its 14A node roadmap discussion. Intel has made no Commission filing referencing the project.

The July 22 statements. REPORTED, from Tesla’s second quarter earnings call. The merger answer is rendered by Reuters as “there’s more and more overlap” and combining companies “can’t” be discussed on an earnings call and must follow “the appropriate process”; one transcript rendering carries the additional clause “especially with the Terafab, that’s really going to be a gigantic project,” recorded here as context and to be confirmed against the primary transcript at print. Tesla’s general counsel, Brandon Ehrhart, followed with the “great partner” and “numerous beneficial transactions” characterizations, per Reuters. The statement that the chip project is necessary to scaling Optimus is from the same call, oral, and appears in no Tesla filing.

The commerce. FILED, Tesla 10-Q, related party note, verbatim in relevant part: $318 million of revenue and $242 million of cost of revenue from SpaceX’s purchase of Megapack products in the quarter, $405 million and $307 million for the six months, in the ordinary course under the Related Person Transactions Policy, with other related party transactions immaterial, and 2025 comparatives immaterial. The first quarter figures are the six month figures less the second quarter. The buyer’s side attribution, “the rapid expansion of our terrestrial data centers,” is FILED in SpaceX’s prospectus, in identical words in each period discussed.

Governance, and the precedent. SpaceX’s structure is FILED in its prospectus and read in The Three Layer Cake: two classes at one and ten votes, about 85 percent of voting power on about 42 percent of the economics, controlled company exemptions, and a 67 percent shareholder proposal threshold. The controller’s Tesla ownership is a matter of Tesla’s proxy record. The SolarCity history is a matter of court record: the 2016 acquisition, with the controller on both sides, was litigated under the entire fairness standard through a 2022 Court of Chancery decision and a 2023 affirmance. It is cited as the one documented local instance of the process invoked on July 22, and for nothing else.

The venue. The Texas Business Court opened September 1, 2024, created by the Texas legislature in 2023; the 2025 amendments to the Business Organizations Code (SB 29) codified presumptions favoring directors and officers and authorized derivative ownership thresholds of up to 3 percent (coverage: Sidley’s Enhanced Scrutiny, May 2025). Tesla’s Texas reincorporation was approved by shareholders in June 2024 and is FILED in its proxy and subsequent charter documents; its adoption of the 3 percent derivative threshold is carried in its governing documents and was the subject of a filed exempt solicitation by the New York State Comptroller (accession 0001214659-25-015361). SpaceX’s Texas incorporation is FILED in the S-1; the mandatory arbitration characterization of its shareholder claim provisions is REPORTED from governance commentary and should be confirmed verbatim against the prospectus’s forum and arbitration provisions at print. The Nasdaq and Nasdaq Texas listing is FILED on the 424B4 cover. The January 2024 Chancery decision and the 2016 to 2023 SolarCity litigation are matters of court record. The sequence is stated as sequence; no purpose is attributed to any step.

The one way flow. The absence of any SpaceX position in Tesla stock is an absence in the record read here: no Schedule 13D or 13G by SpaceX with respect to Tesla exists, and the S-1’s balance sheet and investment disclosures, read in full in earlier entries, carry no such position. Absence of evidence in filings read is stated as exactly that.

The compensation plans. FILED in the respective companies’ 2025 award documents and read in The $8.5 Trillion Grease Gun (May 1, 2026) and The $10.25 Trillion Hypothesis. The Tesla award runs to 423.7 million shares across tranches tied to market capitalization and operational milestones, and its change in control provisions disregard the operational milestones, with the transaction price determining the market capitalization achieved. The SpaceX award comprises 200 million Class B shares with additional tranches tied to orbital data center milestones. This piece cites the mechanisms only. The May piece’s transaction scenario was a hypothetical at pre offering marks and is not repriced or endorsed here; no intention is attributed to any party.

The pile, and the spread. Tesla’s cash figures are FILED as above, and the SpaceX stake is not among them: the policy note states the equity investment is presented within other non-current assets, so the $43.5 billion excludes it. The lockup characterization in the text reads the offering’s general lockup terms, FILED in the final prospectus, against the March 2026 origin of Tesla’s shares; Tesla makes no specific disclosure of its position’s lockup status, and the final prospectus’s lockup schedule should be confirmed at print. SpaceX’s $25 billion of June notes at a weighted average coupon near 5.855 percent, and the $25 billion minimum cash commitment, are FILED (8-K accession 0001628280-26-044955) and were read in Cursor Stock and the Investment-Grade Refinancing. The juxtaposition of one company’s cash yield with the other company’s coupon is an observation about two filed records and carries no suggestion of any transaction between them.

The speed doctrine. FILED. “We believe speed is a competitive advantage... we continue to prioritize execution speed, capacity expansion, and technological leadership over near-term margin optimization,” SpaceX S-1 and final prospectus (0001628280-26-036936; 0001628280-26-042639), fragment-confirmed against EDGAR full text at the prior entry’s print check on July 26. Its reading as the plan’s engine, and the governance structure as the price of the speed, is in SpaceX, Adding It Up: The Cost of Not Beginning.

What consolidation would erase. The elimination of intercompany revenue and of a parent’s mark on a subsidiary in consolidation are descriptions of the ordinary operation of the accounting framework. The segment reporting discussion describes ASC 280’s management approach, under which reportable segments and their disclosed measures follow what the chief operating decision maker regularly reviews. These are descriptions of the framework, not legal conclusions, and no view is expressed on what any combined entity would or should disclose.

What arrives next, and how it lands. Written before any of it, whatever it is. A definitive Terafab agreement would file as a material definitive agreement at both companies, converting the framework’s undetermined projects into determined ones, and would date this piece’s account of the framework to the day before it arrived. An announced combination would begin the process the July 22 answer named; nothing here predicted one, and the ladder above is a list of filings, not a forecast. A denial of any combination changes no present-tense instrument described here. The Texas record’s own dates are arrivals too: the August 2 payment and the dated investment commitments are as legible passing unmet as filings are landing; the payment date falls on a Sunday, so a Monday landing is the calendar working. SpaceX proceeding alone under its already disclosed plan generates no Commission filing, per the previous entry’s own reading of that silence, and arrives in the Texas record instead. And on the mark: a recovery in the SpaceX quote by the September 30 measurement date prints no loss and proves the sensitivity just the same; the claim is the mechanism, not the mark, and it runs in both directions by construction. In every case, the sources of this piece predate it, each is public and dated, and the earliest is this series’ own entry of May 1. A piece built from dated public documents will sometimes publish close to the record’s next document. Proximity is not foresight, and none is claimed.

Analysis: Cape Fear Advisors.

The mark, and what has been filed since

On June 12, 2026, SpaceX sold shares to the public at $135, the largest offering in history, and the market set the company near $1.79 trillion. Six weeks later it closed at $115.07 on July 24, about $1.53 trillion on roughly 13.26 billion shares, after a high close of $201.80 on June 16. (1) Headlines call the move a slump. Marked as a position, it is more particular. The entry is $135 and the plan behind it, and the question is not where the tape goes next but what has arrived since to justify the entry.

The record filed with the Commission since the debut is short. The acquisition of Cursor, disclosed before the offering and settled in stock, which changes the share count and not the cash need. A $25 billion bond, which refinanced and extended existing debt without adding capital the business can deploy. Investment-grade ratings from all three agencies on June 18. On the chip project, nothing: no material definitive agreement, no amendment, no completion. Part of the markdown is weather, because over the same six weeks the broad market softened and the AI trade cooled, so some of the decline belongs to everything and not to this company. What this piece reads is not the cause of any one day but the level across the stretch. (17, 18)

Sixteen of the twenty-eight sessions since the debut moved more than three percent, and seven moved more than five, from a first counted session up 19.5 percent on June 15 to a single session down 16.4 percent the day the bonds priced. Wide sessions are ordinary after an offering this size, and part of that dispersion is a market finding a clearing level for a float that did not exist in June. The amplitude has come down from those first-week extremes. The frequency has not: the last full week in the series moved 3.3, 3.1, 6.7, 2.6 and 2.7 percent, three of five sessions above three percent, six weeks in. Whatever is being argued about has not been settled by trading. (1)

How much of that belongs to the market is answerable, and the answer is most of it does not. From the June 12 close to the July 24 close the stock came down 28.6 percent. Over the identical window the Nasdaq Composite came down 3.5 percent, the Nasdaq-100 tracker 5.1 percent, the S&P 500 three tenths of one percent, and the Dow finished higher. The nearest cohort available in a listed instrument, the semiconductor tracker, came down 9.5 percent, and that stretch did contain a real derating of the AI capital cycle: a foundry’s capital-spending guide on July 16, and a July 23 session in which one large platform fell about six percent on a capital-expenditure guide of $195 to $205 billion. So a cohort move exists, it was narrow rather than market-wide, and at its worst it is about a third of this one. Roughly two thirds of the decline has no cohort standing under it. That does not name a cause, and nothing here names one. It marks how much of the distance a general explanation reaches, and where it stops. (2)

The other ready explanation is company news, and the filed record set out above is the whole of it. None of those three items is a repricing event of that size, and the one that carries a direction carries it upward. A stock-settled acquisition moves the share count and not the cash need. A refinancing adds interest and not deployable capital. Three investment-grade ratings are favorable if they are anything. So the window holds no arriving fact large enough to account for the move, and on the one project the price was carrying it holds no filing at all. What is left standing in the record is six weeks passing against a date the issuer set for itself. This piece does not say the market priced that, because from outside no one can say which disagreement moved on which day. It says the record offers nothing else, and a residual is where a market keeps what it has not been given a way to itemize. (4)

They set the date

The schedule belongs to the issuer. The prospectus states that the company expects to begin deploying its orbital AI compute satellites as early as 2028, repeats that date in the forward-looking statements, and the roadshow materials filed with the offering carry the same line. The date was not buried in a risk factor. It was shown to buyers as part of the offer. (7)

The prospectus then names what stands between the company and the date. “Our ability to achieve orbital AI at scale depends on our ability to access a sufficient number of AI chips, significantly more than are currently available to us.” It names one answer, and the consequence of that answer failing. “While we expect to construct Terafab to address such supply constraints, Terafab may not be successful, in which case we may not have other sources of sufficient AI chips to meet our orbital AI compute demands.” The back half of that sentence is the company stating, in its own risk factor, that it has no second source. (7)

That phrase invites a reading it will not bear. The risk factor is about the sufficiency of supply for the company’s own demand, enough chips, in time, at the volume a constellation requires. It is not about whose silicon flies. On that second question the filed record says something close to the opposite of exclusivity: “We intend with our SpaceX AI satellites to allow people to put whatever GPU or TPU they want.” A customer may bring its own accelerator. The company may still be unable to obtain enough of its own. The two statements sit in different places and neither one answers the other. (8)

The controller said the same thing about supply in a filed interview, and said it as necessity. Asked by Jamie Dimon what compelled the chip build now, Musk named the constraint first. “What we see as a limiting factor is being able to make chips, both logic, memory, and packaging.” Then the state of the ground. “There’s not a single high volume computer memory fab in America right now. Zero.” The plant under construction in Idaho does not reach volume production until 2028 on his account, the New York projects later still, and then the arithmetic that decides it: “take the best case assumptions of the memory makers and the logic makers, it is not enough to meet the demand that is anticipated.” Then the conclusion, in three short sentences. “And that’s why we need to do Terra Fab. It seems essential. Otherwise we will not, there will not be enough chips.” The prospectus deals in may and will. This is a need, and a need cannot be deferred without cost. (8)

The paperwork behind the need is thinner than the language. What exists for the facility the company expects to be the world’s largest is “a general framework,” with financial terms, intellectual property rights, and the ultimate term of the collaboration all, in the prospectus’s words, not finalized. And there is the sentence the staff’s comments produced, which the prospectus repeats in four separate places: “Any specific projects undertaken pursuant to this framework will be subject to separate negotiations and agreements (including any development timelines, milestones and capital expenditures) and have not yet been determined.” Not the timeline undetermined. The projects. (12)

The record that is not filed

Beside that sits a second record, and it is not thin at all. It is in a Texas county clerk’s office, in two school districts, and with the state Comptroller, and this series read it in full in The Terafab Record. Between June 3 and July 21 that record filled in. A subsidiary named TeraFab AI, LLC has eight applications on file under the state’s Jobs, Energy, Technology and Innovation program, four with Anderson-Shiro CISD and four with Iola ISD, describing investment of up to $119 billion in phases and seeking roughly $1.66 billion of maintenance-and-operations tax benefit. A county judge signed a reinvestment zone. A deputy comptroller signed a recommendation on all eight. A county agreement took effect on June 3 over the county judge’s signature and was signed by the company’s chief financial officer on June 22, committing at least $5 billion of investment in Grimes County by 2030 and at least 1,800 full-time jobs by 2035, with $10 million payable and non-refundable by August 2, and a right to terminate at any time and for any reason on thirty days’ written notice, at a cost that stays small until a date the agreement itself names. Both school boards approved on July 14 by five-to-two votes. On July 21 the Governor’s positive determination was confirmed by the Comptroller’s office, though not by his own, and no verification letters have posted. Local coverage of the Iola vote reports a first construction phase intended before year end. (9)

None of that is in a filing. Not one line of it appears in an 8-K, and the two words that would carry it, a definitive agreement and a completion, are the two triggers the rules make mandatory. So the question of what to make of it has an answer already given, and given by the issuer. (10)

The answer is that all of it is ordinary course and none of it is material. That is not a reading imposed from outside. It is the company’s own determination, made the way determinations of this kind are made, by filing nothing, and the sizes agree with it. Ten million dollars is about one tenth of one percent of a single quarter’s capital spending at this company. Five billion by 2030 is less than a quarter of what the segment spent in 2025 alone. Against a $1.5 trillion market capitalization these are rounding. A commitment that can be exited on thirty days’ notice for a sum in the tens of millions is not a material definitive agreement in the sense the rules mean, and a permit, a zone, a school board vote and a governor’s determination are the ordinary machinery of putting a plant somewhere. The company judged all of it immaterial, and on the numbers the judgment is defensible. (10)

What follows from that runs the opposite way from the reading a holder might reach for. The absence of filings is not evidence that nothing is happening. It is evidence that what is happening is what the company already told the market it would do. The prospectus says Terafab will be constructed. That statement makes the attempt part of the ordinary course of the business, and an abandonment of the attempt, not its pursuit, would be the change large enough to require saying so. Silence on the attempt is the baseline holding. Progress against a plan already disclosed is exactly the thing that generates no news.

That settles what this piece is adding up, and it is not the cost of nothing happening. Something is plainly happening in Grimes County. The cost is the distance between what is happening there and what the filed record carries, because the filed record is the one the price was set on.

What ordinary course carries

The baseline carries the attempt. It does not carry the date, and the difference between those two is where the money sits.

The company has an account of its pacing, given by its chief financial officer in a filed interview. “We really take an approach from a capital allocation similar to what you would do in a just in time model.” On that doctrine capital arrives when it is needed and not before, which is discipline rather than delay, and a holder demanding to see spending early is asking the company to be worse at capital allocation than it says it is. (15)

The filing supplies the other half of that picture with the company’s own accounting. Capital expenditure in the AI segment ran $463 million in 2023, $5.6 billion in 2024, $12.7 billion in 2025, and $7.7 billion in the first quarter of 2026 alone, against $2.6 billion in the same quarter a year before. This is not a company that is slow to spend or short of the means to spend. And the prospectus says, in the same words every time it explains the increase, where the money went: “the rapid expansion of our terrestrial data centers, including the development, construction, and equipping of new facilities and supporting infrastructure.” Terrestrial. The capital is moving at scale and at speed, and the filed attribution puts it on the ground rather than in a fab. The state applications say the same thing in their own way, scheduling the heavy Terafab spending into 2028 and beyond. (13)

Two records of one project ran in parallel across six weeks at very different levels of specificity. One of them names phases, dollars, acreage, water, a payment date and an exit. The other says the projects have not yet been determined. Both are accurate to their own purpose, and neither is a failure of the other. What is true, and what a holder carries, is that the record with the specificity is not the record the securities laws deliver, and the holder who reads only what is filed sees the less specific of the two. The distance between them is not a disclosure defect. It is a measurement problem, and measurement problems have prices.

One model a reader can open

Of the four families of valuation, intrinsic, relative, asset-based and contingent claims, three go dark against this company: there is no true comparable for a business that is part launch, part satellite network, part data center and part chip ambition, so any peer group is a choice that makes the answer; there is no acquirer at a trillion and a half, founder-controlled and defense-adjacent, so the takeover premium that floors most companies is unavailable; and replacement cost is a different kind of floor from the one a franchise is priced on. Only real options has a slot for the question the market is arguing about.

Inside that one surviving family the practitioners disagree by a factor of six, measured from the lowest posted model to the highest bull case. Aswath Damodaran, who posts his model publicly, values the equity near $1.3 trillion, about $98 a share, and puts a band of $1.25 to $1.35 trillion around it, which on the prospectus count is about $94 to $102. Morgan Stanley is at $300, with a bull case at $600. JPMorgan is Overweight at $225. HSBC joined on July 24 with the lowest published mark, Hold at $115, on a sum of the parts to which coverage reports it added a doubling for the founder’s record, and it named the reason in the terms this piece has been using: investors, it wrote, are “already pricing in technologies that remain unproven, including orbital data centres, AI compute infrastructure in space and semiconductor manufacturing through its Terafab project.” Its own blue sky is $293. The published range now runs $115 to $600, and the low end doubles the businesses for the controller and still lands twenty dollars below the price the company set. The dispersion is a fact about the asset rather than noise to be averaged away: a company whose value almost no one agrees on, because almost all of it rests on things that have not happened yet. (14)

On Damodaran’s posted spreadsheet the operating business is worth $1.224 trillion and the equity $1.301 trillion, $97.83 a share, at a cost of capital of about 8.4 percent. Inside it sits a revenue line for other businesses, the large markets he thinks the company may enter and does not name. With that one line set to zero and nothing else changed, his own model returns $80.36 a share. The difference, $17.47, is what his model pays for everything the company is not yet in. He does not name that line to any one project, and the chip project is one of the things standing inside it rather than the whole of it. The price divides into three layers, all in his numbers and the tape: about $80 of businesses at their expected performance, about $17 of his own expansion-options line, and above both of those a residual, the part of the price his model does not reach. The first two layers are his and were not re-run for either date. Only the third one moved. It was $37.17 at $135 and it was $17.24 on July 24, so the entire $19.93 between the two marks sits in the layer above everything his model is willing to count. (3, 4, 5)

That $80 figure is a floor inside his model rather than a floor under the company, and it is not a pessimistic one: it still carries his full expectation for the AI business, revenue climbing from about $3 billion to a $160 billion target by 2036, which is itself an attempt. He would not accept the word optimistic for any of it, and the objection belongs in the record. He describes his own estimates as conservative, on the ground that he leaves out space travel and expanded business opportunities, and he defends a cost of capital close to the median for all US companies because most of the risk here is specific to the company, thins out in a diversified portfolio, and “cuts in both directions (upside and downside).” One input changes here, his own, and his model speaks. What survives the objection is that the floor sits well below the price, and the distance between them is what the tape has been arguing about. (3)

That distance is measurable at three marks. At $135, about $55 a share sat above the $80 floor, roughly 40 percent of the price, and closer to half given that the floor already assumes the aggressive AI case. At the June 16 high close of $201.80 the distance was $121. On July 24 it was $35. Damodaran’s own expansion line is a modest version of the same aspiration, $17.47, and he was candid about what it is, calling it in the earlier version of this valuation “a crude attempt to capture this part of the story.” Against that line the market paid about seven times at the high close, three times at the offering, and twice at the most recent close. Those multiples are a ruler and not an attribution. What sits above his fair value is a residual, and a residual absorbs every disagreement a buyer might have with him: his cost of capital, his revenue path, his terminal assumptions, and the attempt among them. That is not the same as calling it opaque. He names the largest of those disagreements himself, hands over the lever for each, and says plainly that “disagreements about market size and profitability across investors, especially in young companies, are natural and healthy.” What cannot be done from outside is to say which of them moved on any given day. What can be said is that the market assesses the whole of that quantity and prints its assessment at the close of every session, and that between two of those closes it came down by $19.93. (3, 4)

That last figure carries a caution that is best said in the open. Because his model was not re-run between the two dates, the fall in the third layer is arithmetically the same event as the fall in the price: $135 less $115.07 is $19.93, and no decomposition was needed to reach it. Any fixed valuation, his or anyone’s, subtracts out and leaves the same number; his low end gives premiums of $41.26 and $21.33, his high end $33.28 and $13.35, and the difference is $19.93 in all three. So the layers are not a measurement of the move. They are a measurement of the level: how much of the price on a given day stands above what a model with its working shown will pay, and how many multiples of that model’s own line for the not-yet-begun the market was carrying. At $135 it was three times that line. At the June high it was seven. On July 24 it was two. That is the reading, and it is the whole of it. (3)

He wrote something in April, before the prospectus existed, that sits against the tape. He ran that earlier valuation ten thousand times, took the median of the runs at $1.29 trillion, and said of the pricing rumored at the time that “a $1.75 trillion or even a $2 trillion pricing falls in the range of the distribution, though with little or no upside left for an investor paying that price.” The company priced in June at $135, which on the count used here is $1.79 trillion. That is the sequence. No cause is claimed from it, and he claimed none himself. (3)

What this piece does with that is reconcile the news to the changes in the tape, and the limit of the exercise should be said out loud: a move can come from news, from volatility, or from a reassessment of the residual, and from outside the tape those three cannot be separated. The one outside reading of the same stretch that puts a reason on paper is HSBC’s, initiating at the bottom of the published range on July 24, and it lands on the last two, pricing for technologies that remain unproven. It is a bank’s note reaching the record through coverage, and it is offered here as that and nothing more. (14)

Whose numbers, then. The controller set the offering price at $135, and a buyer that morning bought whatever the not-yet-begun was carrying at whatever he had set it. No other price was on offer. What is left to an outsider afterward is the choice of a yardstick, and this one is used because it is the least favorable to the case made here. His floor leaves the standing businesses $80.36 of the $135, about three fifths of it, and about seven tenths of the July 24 close. HSBC’s implies less: coverage reports its $115 rests on a sum of the parts the bank doubled for the founder’s record, which puts that sum near $58, somewhat over two fifths of the offering price. The rest of the published marks show no working at all. A range could be drawn across them, and it is not drawn here, because the one figure a reader can open and test is also the one that hands the standing businesses the largest share of the price. (6, 14)

He does not argue for anything lower, and it would be a misuse of him to suggest otherwise. He calls his own estimates conservative, he says the outliers are likelier on the upside than the downside, he declined the offering as too richly priced for his tastes rather than calling it overvalued, he will not sell it short, and he wrote in April that the company was one large correction away from being fairly priced or cheap and that he would be a buyer if that came. The model is his. Every use made of it here is this piece’s own. (6)

Lower numbers do exist inside that model, and reaching them requires no disagreement with him at all. His probability-of-failure input sits at zero, which is where the workbook starts. Work published here on June 7 measured the lever on that input: on his own convention that a failure returns about half of fair value, each percentage point takes roughly $6.1 billion off the operating assets, about 49 cents a share. His own published precedents, from his prior valuations of young companies, run 5 percent for Uber, 10 percent for Lyft, for Airbnb and for Zomato, and 20 percent for WeWork. At 10 percent his $97.83 becomes about $92.96, within a dollar of the $93.74 low end of his own June band. At 20 percent it is about $88.10. Both still stand above the $80.36 floor used here, so the shape of the reading does not change. Neither is offered as a correction of him, and neither is the number this piece is about: that switch is firm-wide and binary, and a schedule is not what it measures. (3, 5)

What the company can do now to earn the premium is a short list. On the AI compute side the best available outcome is to deliver, in full, the two contracts registered as material, about $1.25 billion a month from Anthropic and $920 million from Google. Those are already in every model, so delivering them meets the number rather than beats it, and both cancel on ninety days’ notice, which makes even full delivery revocable each quarter. The outcomes on compute are capped at zero on the upside and open on the downside. The only positive surprise left lives above the contracts, in the build. (15)

No slot in the model, no slot in the rules

Underneath the forecast is the assumption that matters most, and it arrives by default rather than by choice. The model carries a switch for the probability that the firm does not survive, and it is left off, which is where it starts; the note beside that input observes that young companies fail when they have trouble raising cash. Off is a survival assumption, and survival of a value built on growth is not passive. It assumes the company attempts, and it assumes the attempt lands roughly when the plan says. Every model books the benefit of the attempt, because the attempt is the plan and most of the value. What none of them carries is a state in which the attempt proceeds and the calendar moves, and the failure switch would not carry it either: it is firm-wide and it is binary, and it has nothing to say about a date. Booked, uncosted, and unconditioned. (5)

The reporting rules have the same hole in the same place. What obliges a current report is an event: entry into a material definitive agreement, its termination, the completion of an acquisition or disposition. Two further items exist for anything else, one for selective-disclosure cures and one for whatever the registrant chooses to report, and both are voluntary. Beyond that, a forward-looking statement generally carries no duty to update as circumstances change. So the rules capture things that happen, on the day they happen, at a size that clears a threshold. A date that slides carries no event, no counterparty, and no dollar figure, which leaves it outside every mandatory trigger there is.

That was tested on July 22. On Tesla’s second-quarter earnings call the controller of both companies, asked about the chip project, said that a location is expected to be announced soon and that details would come with the product launch rather than an earnings call. Four days later no announcement has followed. What that sentence is, and where it lives, decides how far it travels. It is oral, it is a timing statement about the project the prospectus puts between the company and the date, and it was made at a related party, on a call belonging to a different issuer, whose own filings have not used the word Terafab since March. A SpaceX holder who reads only SpaceX filings did not receive it. Nothing about that is irregular, and nothing about it was required to be filed. The rules have no slot for soon, in the same way and for the same reason that the model has no slot for later. (11)

What time costs

An option with no stated expiry does not obviously decay, which makes it fair to ask why the premium should erode rather than wait patiently. Beyond the 2028 schedule, the record supplies three more clocks. The compute contracts run to 2029, so the window in which a differentiated build could compound inside contracted demand narrows by a quarter every quarter. The customers are standing up their own capacity across 2027, the demand this series has watched approach. And the sharpest and most filed of the three is the rent. At the debut the company sold $25 billion of notes across five tranches at a weighted average coupon of about 5.86 percent, refinancing and extending the bridge loan it took in March, about $20 billion at roughly 4.6 percent. That is roughly $1.46 billion a year in interest, about half a billion more than the bridge it replaced, the first payment due January 15, 2027, running from June whether or not the schedule holds, and locked out as far as 2056. It prints in the next quarterly filing as interest expense beside whatever the capital line shows. Separately, the company committed in its filings to hold $25 billion in cash, a commitment the ratings rest on, capital it cannot deploy. Time is not free to this company. It has a coupon. (18)

The market is marking the wait in two places. The equity premium above the floor has narrowed from about $121 at the June high to about $35. The bonds, the June investment-grade debut, slipped below issue within days, about $305 million of paper losses, with the ten-year tranche’s yield climbing toward 6 percent and its spread above 160 basis points, junk-like against the grade on the label. The credit market’s stated reason is the reading in its own words: attention has moved from the growth narrative to cash flow and execution, against a first-quarter loss of $4.28 billion. The debt raised to earn a rating is being second-guessed by the market that trades it, and it added no capital the business can spend. Two witnesses, equity and credit, to one uncertainty. (18, 19)

The one who can begin

The schedule has an owner, and the structure that gives it one was part of what the shareholders bought. SpaceX is a controlled company. There are two classes of stock, one vote a share and ten, the founder controlling about 85 percent of the vote on about 42 percent of the economics and the public about 15 percent of the vote on about 60 percent. It is exempt from the requirement that a majority of its board be independent, the controller elects and removes the directors who hold control, and a shareholder proposal requires 67 percent of the voting power, which the vote structure places out of reach. The plan calls speed the advantage, and this structure was the price of the speed, not incidental to it. The right to set the pace of the attempt therefore sits with the controller, and the holders who bought it have no mechanism to affect it. (16)

The exposures run together on the same clock. The premium narrows as the window narrows, execution risk runs beside it, and financing risk sits under both, a company that has turned from returning capital to raising it against a funding gap larger than the offering. The financing has a reflexive edge, because the premium is also the currency: acquisitions have been paid for in the company’s own stock, xAI and Cursor both, so the mark that comes down is the coin the build would be bought with, and its decay raises the price, in dilution, of the expansion it was meant to fund. What helped the plan at $135 works against it at $115, without anyone deciding it should.

The other way an option ends

An option can also end without decaying. The holders’ claim on the attempt has no stated expiry, which is not the same as having none. The party who alone sets the pace of the build can also do other things, and some of them would settle the question by removing it rather than answering it. Asked on the same July 22 call about combining the two companies, the controller answered that it “has got to be done with the appropriate process.” Nothing on the subject is filed, and neither company has said a combination is contemplated. The merger question is a potential action and belongs in a different box from the timing statement made on the same call: one would change what the option is written on, the other changes when it pays. What the question illustrates is structural, and it reaches back to the vote: a holder waiting for an attempt waits inside a period whose terms can be changed by the same party, through acts the holders have no vote to affect. The defense of waiting is that flexibility has value. It assumes the flexibility survives the wait.

A shareholder bought two things, the businesses and a promise to attempt more, and the second is an option. The cost is not that nothing has begun. It is that the beginning which exists is not the beginning the filed record carries, and the one that would carry it, capital deployed at the scale the plan describes, is scheduled for years the coupon has already started running toward. Between those two lies uncertainty about timing that no model books and no rule requires anyone to report. In theory that uncertainty can raise what an option is worth. What it has plainly raised is the volatility of the stock, and the equity and the bond tape are both marking it. Good news could lift the premium again, and the longer the interval, the more of the price rests on the premium alone and the less on anything filed. What a holder is paying for, and paying to hold, is the time value of a promise to try, not even to succeed.

Related, on the Quality of Cash shelf: The Quality of Cash: What Has to Happen Next (the anchor); SpaceX, Adding It Up: The $235 Billion Cash Gap; The Terafab Record; The Ninety-Day Annuity; The Probability of Failure; The Three Layer Cake: SpaceX’s Governance Structure; Cursor Stock and the Investment-Grade Refinancing; SpaceX Confuses Currency for Capital.

Standing disclosure

Anthropic is the developer of Claude, which is used in preparing this research, and Anthropic is also a compute counterparty to SpaceX, the customer whose contract is read here. That nearness cannot be fully checked away, which is why no claim here rests on trust in the tool: every figure carries a public source, and the record grades the rest. Others named have ties to Anthropic: Google, the second compute counterparty, and its parent Alphabet, which holds a large position in Anthropic; and companies not named here, among the chip and cloud suppliers, may hold positions or supply relationships that bear on the filers discussed, which is part of why every piece is re-checked for bias, ground facts, and filings rather than read against a fixed list. Figures are quoted from the filers and from named analysts without characterization, and the same standard of reading is applied to every party named.

NOTES

How to read the labels. Every load-bearing figure below carries a source and a status. FILED means a document filed with the Commission and cited by accession, which any reader can pull. REPORTED means the public record outside the Commission’s files: county and state documents, local coverage, market data, and analyst work reached through coverage or posted by its author. The distance between those two words is what this piece is about, so they are used strictly and nothing is left unlabeled.

(1) Prices and size. REPORTED market data, with the offering itself FILED. IPO priced $135 on June 12, 2026; the first close was $161.11, 19.3 percent above the offering price. Closing prices only are used throughout, and intraday prints are excluded: the highest close since the debut is $201.80 on June 16, 2026, and the most recent close pinned here is $115.07 on July 24, 2026, a market capitalization near $1.53 trillion. The share count used for both capitalizations in the text is 13,257.5 million: Damodaran’s 12,535.3 million basic shares, which are the shares outstanding before the offering and the denominator his per-share figure divides by, plus the 638,888,888 Class A shares sold, plus the 83,333,333 over-allotment shares. At $135.00 that is $1.790 trillion and at $115.07 it is $1.526 trillion. Excluding the over-allotment the count is 13,174.2 million and the two figures are $1.779 trillion and $1.516 trillion, which is within about a percent at each mark; the fuller count is used because the over-allotment was in fact exercised. The session-move counts, sixteen of twenty-eight sessions above three percent and seven above five, are computed from the same reported daily closes, June 15 through July 24, close over prior close, so the first move inside the count is June 15’s 19.5 percent over the June 12 close rather than the debut’s 19.3 percent over the offering price; June 19 and July 3 were market holidays. No session in the twenty-eight sits at exactly three or exactly five percent, so the strict reading and the inclusive one return the same counts. This piece publishes on Sunday, July 26, 2026, so the July 24 close is the last session before publication and no session intervenes between the pinned price and the print. FILED: SpaceX Form 8-K reporting the IPO close (accession 0001628280-26-043288) records 638,888,888 Class A shares sold at $135.00. The base offering raised about $75 billion; the underwriters’ over-allotment (the green shoe) of 83,333,333 shares was exercised in full, about $11.25 billion more, taking the gross to about $86.25 billion. Damodaran’s model, written at pricing, uses the $75 billion base, and his per-share figure is not merely little affected by the over-allotment but exactly invariant to it: the per-share result is equity less the offering proceeds, divided by the pre-offering basic share count, so proceeds and the shares issued for them cancel and the size of the shoe drops out.

(2) The cohort. REPORTED, all of it, and none of it FILED. Closing levels on June 12 and July 24, 2026: SPCX $161.11 and $115.07, down 28.58 percent; Nasdaq Composite 25,888.84 and 24,975.82, down 3.53 percent; the Nasdaq-100 tracker QQQ $721.34 and $684.23, down 5.14 percent; the S&P 500 7,431.46 and 7,411.98, down 0.26 percent; the VanEck semiconductor tracker SMH $619.96 and $561.19, down 9.48 percent; the Dow Jones Industrial Average 51,202.26 and 51,947.25, up 1.45 percent, recorded because a rising index is the harder fact for a market-wide reading. Measured on closes throughout, on the same convention used for the stock. The Composite’s worst closing drawdown inside the window is 6.40 percent, from a June 15 peak of 26,683.94 to the July 24 close. The stock’s decline is about three times the deepest cohort proxy and about eight times the Composite. Window events referenced in the text are reported: TSMC’s capital-expenditure guidance on July 16, 2026; the July 23, 2026 session in which Alphabet fell about 5.9 percent on 2027 capital-expenditure guidance of $195 to $205 billion and the Composite fell 2.15 percent. Also inside the window and not separated out here: a June 16-17 FOMC meeting at which forward guidance was withdrawn, crude oil up about 6 percent, the ten-year note near 4.675 percent, and newly implemented tariffs of 10 to 12.5 percent. No causal claim is made from any of it. The comparison is used only to bound how much of the move a general market explanation can carry.

(3) The model. Neither FILED nor REPORTED, and the distinction is the reason this piece uses it: a named practitioner published his valuation with the spreadsheet attached, so any reader can open it, change the one input changed here, and re-run it. Aswath Damodaran, “Revisiting the SpaceX Valuation: A Post-Prospectus Update”, June 2026, and his posted spreadsheet (SpaceX2026IPOUpdated.xlsx); the earlier valuation it updates is “To a Trillion(s) Dollars and Beyond: A SpaceX IPO Odyssey”, April 2026. Value of operating assets $1,224,448 million; value of equity $1,301,299 million; value per share $97.83; basic shares 12,535.3 million; cost of capital 8.37 percent initial and 8.25 percent terminal; probability of failure input set to zero. Setting his “Revenues (Other) in 2036” input to zero, with every other input and convention unchanged, recomputes the equity at $1,082,278 million and the per share at $80.36; the difference, $17.47 a share, is the value his model assigns to the expansion options. The per-share figure reduces to (equity minus IPO proceeds) divided by basic shares. The three distances above the floor cited in the text are $54.64 at the $135 offering price, $121.44 at the June 16 close, and $34.71 at the July 24 close, which are 3.1, 7.0, and 2.0 times his expansion-option line. Measured instead against his fair value of $97.83, the market premium was $37.17 at the offering and $17.24 at the July 24 close, a decline of $19.93. His floor and his expansion-option line are outputs of a model rather than marks in the tape, and the model was not re-run for either date: its riskfree input is 4.56 percent, dated June 1, 2026, and a 25 basis point change in that one input moves the per-share result by roughly $2.15 to $2.57. So neither of the first two layers differs between those two dates, and the whole of the move sits in the third by construction. That is what the two columns of the layers exhibit show. Because a model value that does not change between two dates subtracts out of both premiums, the $19.93 is invariant to which of his figures is used and is arithmetically identical to the change in the price itself: against his band of $1.25 trillion to $1.35 trillion for the equity, which reduces to about $93.74 and $101.72 a share on the same convention, the premium runs $41.26 and $21.33 at the low end and $33.28 and $13.35 at the high end, a fall of $19.93 in each case. The decomposition is therefore used in this piece for the level and not for the move. In the earlier April valuation he reports a base of $1.22 trillion and a median of $1.29 trillion across ten thousand simulation runs; that post states no other percentile, gives no share count and no per-share figure, so nothing per-share is taken from it here. The probability-of-failure arithmetic cited in the text is this piece’s own, computed on his posted workbook and on the measurement published in SpaceX, Adding It Up: The Probability of Failure, June 7, 2026: with a failure modeled to return about half of fair value, each percentage point of failure probability removes about $6.1 billion of operating asset value, which on 12,535.3 million basic shares is about $0.487 a share, so 10 percent gives about $92.96 and 20 percent about $88.10 against his $97.83. The precedents quoted are his own from prior valuations, 5 percent for Uber, 10 percent for Lyft, Airbnb and Zomato, 20 percent for WeWork. He does not apply any of them to this company, the input in the posted workbook is at zero, and no claim is made here that he intended otherwise.

(4) What the third layer is, and is not. It is a residual, and it is placed in that bucket here knowingly. It is the price less his fair value, so it absorbs every disagreement a buyer might have with his forecasts, not the chip project alone: a 50 basis point cut in his terminal cost of capital by itself moves about $12.00, against an expansion line of $17.47. Part of what a buyer pays for the chip project may also sit inside the first layer, in the AI segment his floor still carries in full, so the layers are not clean partitions of the thing they are named for. The decomposition is used here as a scale for reading the size of the move, not as an attribution of the move to any one cause. The market does the attribution, once a day, at the close, and the whole of this piece is an attempt to reconcile a filed record to that number rather than to replace it.

(5) What his expansion line is called, and when it runs. His workbook labels the segment three spokes and his April 2026 post describes it as unnamed large markets, plural. It is not a Terafab line and is not read as one here. It is also not undated: his revenue row for it ramps from zero beginning in 2032. His own option-value sheet is switched off in the posted workbook, so the line is a revenue segment inside the discounted cash flow and is called that throughout rather than an option value. The crude attempt phrase is his, from the April 2026 version of the valuation: “In a crude attempt to capture this part of the story, I will attach an expected revenue in 2036 to these other businesses of $50 billion and an operating margin of 30%”; the June input used here is $100 billion, double the April figure. The characterization of the line as low probability and high payoff is this piece’s summary of his discussion and not his phrasing. What he writes of these lines is that they are “all options that may not be viable at the moment, but if they become viable, could add immense value”, and of one of them that the possibility “may be low, but it does exist.” The June post refers to the “expansion options embedded in each business”, plural and business by business, which is a further reason not to read the line as a single project. The probability-of-failure input discussed in the text is the workbook’s default setting rather than an affirmative choice about this project, and it is a firm-wide distress truncation with no connection to any project’s schedule.

(6) Why this model, and what the others imply. Damodaran’s floor of $80.36 is 59.5 percent of the $135 offering price and 69.8 percent of the $115.07 close on July 24. HSBC’s $115 target is reported through coverage to rest on a sum-of-the-parts valuation that the bank then doubled for the founder’s record; read literally, that puts its sum of the parts near $57.50, which is 42.6 percent of the offering price and half of its own target. Morgan Stanley at $300 with a $600 bull case and JPMorgan at $225 publish no decomposition, so no such share can be computed from either. The doubling is a characterization in coverage and not a figure taken from the note, which is not public, and it is used here only for the comparison stated in the text: among the published marks that show any working, the one used here assigns the standing businesses the largest share of the price. His own posture is recorded so that it is not mistaken for this piece’s: in April he wrote that the company was “one big correction away from being fairly priced or even cheap” and that “If that happens, I will be a buyer”; in June he wrote that at the rumored $1.8 trillion it was “too richly priced for my tastes” while adding that “That does not mean that I will never buy the stock”; and in both he says he would not sell it short. He describes his estimates as conservative and writes that if there are outliers they are likelier on the upside. He nowhere calls the stock overvalued, and this piece does not report him as having done so.

(7) Timing. FILED. The 2028 date is the issuer’s, stated three ways: “We expect to begin deploying our orbital AI compute satellites as early as 2028” appears in the business section, in the AI compute discussion, and in the forward-looking statements list of the S-1 (accessions 0001628280-26-036936 and 0001628280-26-042639); the roadshow materials filed as a free writing prospectus (accession 0001628280-26-040610) carry the same deployment line. The chip constraint and the absence of a second source are quoted from the S-1 risk factors verbatim.

(8) Terafab, and the spelling. The S-1 spells the project “Terafab” as one word. The filed interview transcript renders it “Terra Fab,” two words, throughout; earlier entries in this series noted the same difference. Both spellings appear here as they appear in their sources, and nothing turns on the difference. The interview is a Rule 433 free writing prospectus, a June 4, 2026 video interview of Elon Musk by Jamie Dimon of J.P. Morgan hosted by Mary Erdoes, filed in connection with the registration statement (accession 0001628280-26-041365), with the transcript filed as Exhibit B to that free writing prospectus. The quotations here are FILED, taken from that exhibit, which carries the filer’s own note that the transcript was “slightly edited for clarity.” Quoted in the text: “What we see as a limiting factor is being able to make chips, both logic, memory, and packaging”; “There’s not a single high volume computer memory fab in America right now. Zero”; “take the best case assumptions of the memory makers and the logic makers, it is not enough to meet the demand that is anticipated”; “And that’s why we need to do Terra Fab. It seems essential. Otherwise we will not, there will not be enough chips”; and “We intend with our SpaceX AI satellites to allow people to put whatever GPU or TPU they want.” On the scale of the plant, “we expect that the Terafab is going to be around 100 million square feet, which is 10 times the size of the Tesla Gigafactory Texas” (free writing prospectus, accession 0001628280-26-041761, Exhibit B).

(9) The Texas record. REPORTED throughout and FILED nowhere: these are county, school-district and state records, and local coverage of them, none of which is a document filed with the Commission. Read in full in The Terafab Record. Eight applications under the Texas Jobs, Energy, Technology and Innovation Act were filed by TeraFab AI, LLC, four with Anderson-Shiro CISD and four with Iola ISD, describing phased investment of up to about $119 billion and seeking about $1.66 billion of maintenance-and-operations tax benefit. Grimes County commissioners created a reinvestment zone in early June, and the county released the agreement documents on June 6 (KBTX); a deputy comptroller signed recommendations on all eight applications on June 15 (Comptroller, J0038 Anderson-Shiro recommendation; KBTX, June 18); the county agreement carries an effective date of June 3, 2026, the date the county judge signed, and SpaceX’s chief financial officer signed on June 22, reported June 23 (KBTX). The two dates are separate and both are correct; the payment clock runs from the effective date. The county agreement’s principal terms as reported: at least $5 billion invested in Grimes County by 2030, at least 1,800 full-time jobs by 2035, a $10 million non-refundable payment due August 2, 2026, a Sunday, which KBTX reports as sixty days from the June 3 effective date, and a right to terminate “at any time and for any reason” on 30 days’ written notice, with the cost of exit rising on a schedule the agreement names. Early press rendered the exit cap as $60 million; The Terafab Record works from the posted agreements rather than that characterization, and readers who need the mechanics should read it there. Both school boards approved on July 14 by 5-2 votes (KBTX, July 15; The Texan); Iola coverage reports a first construction phase intended before year end. On July 21 KBTX reported, on confirmation from the Texas Comptroller’s office, that the Governor submitted a positive determination; verification letters have not posted, the confirmation was verbal, and the Governor’s office did not respond to requests. Remaining steps per that reporting: formal execution by the applicant, both districts and the Governor, submission to the Comptroller, and then the first construction phase. Also in the record and not relied on here: a July 2 county request for an Attorney General opinion on withholding certain SpaceX communications from a records request (KBTX), and a July 9 report on incomplete applications and weeks of corrections in the state file (KBTX; KWTX). The two stations are sister outlets and carry the same Terafab reporting, so items below are cited to whichever posted the text version. Sources: KBTX and KWTX (Bryan-College Station and Waco), The Texan, and the posted county and Comptroller documents.

(10) Materiality, and whose judgment it is. The absence is itself the evidence, so it is stated plainly: no document FILED by SpaceX with the Commission carries any part of the Texas record. The characterization of it as ordinary course and immaterial is the issuer’s own determination, evidenced by the absence of a filing, and nothing here asserts that anything was required to be filed and was not. The arithmetic supports the determination: the $10 million payment is about 0.1 percent of the roughly $10.1 billion of capital expenditure in the first quarter of 2026 alone, and the $5 billion investment commitment through 2030 is under a quarter of the $20.7 billion of company-wide capital expenditure recorded for 2025. The relevant current-report triggers are Item 1.01 (entry into a material definitive agreement), Item 1.02 (termination of one), and Item 2.01 (completion of an acquisition or disposition of assets). Items 7.01 and 8.01 exist for Regulation FD disclosure and for other events at the registrant’s option, and are voluntary. A forward-looking statement generally carries no duty to update absent a duty to correct. This paragraph describes the framework, not a legal conclusion about any particular item.

(11) The July 22 statements. REPORTED. Elon Musk on Tesla’s second-quarter 2026 earnings call, July 22, 2026. On the chip project, published transcripts render the line as an expectation to announce a location soon, with details to come at a product launch rather than on an earnings call; transcript providers differ on one word, one rendering “a location” and another “the location,” a difference that bears on whether a site is settled and pending announcement, so the text above paraphrases rather than quotes on that point. Sources reviewed: Investing.com and Alphastreet transcripts and a Not a Tesla App summary. On combining the two companies, the quoted words “has got to be done with the appropriate process” are reported by Fortune (Shawn Tully, July 25, 2026). Tesla’s own second-quarter earnings exhibit and its Form 10-Q contain no use of the word Terafab; Tesla’s filed count of the word remains zero since March 2026. Also said on that call, and not used here: a statement that the chip project is necessary to scaling Optimus production, which is scoped to Tesla’s own demand and not to SpaceX’s orbital deployment, and a description of a separate development fab in Austin, which is Tesla’s facility and Tesla’s spending. No filing on a combination exists, neither company has stated on the record that one is contemplated, and nothing here should be read as predicting one.

(12) The undetermined-projects sentence. FILED. Verified verbatim in the S-1 at four separate places (the summary, the business overview, the strategy section, and the Terafab discussion): “Any specific projects undertaken pursuant to this framework will be subject to separate negotiations and agreements (including any development timelines, milestones and capital expenditures) and have not yet been determined.” The scope is the projects, not only the schedule. The characterization of the arrangement as “a general framework,” with financial terms, intellectual property rights, and the ultimate term of the collaboration not finalized, is the prospectus’s own; the revision arose from the staff’s comments (comment-letter correspondence). Nothing here treats that sentence and the Texas applications as inconsistent: the framework sentence concerns the collaboration with the named partners, and the state filings are conditional, phased, and exitable.

(13) Capital expenditure, by segment. FILED. From the S-1’s MD&A and segment discussion: AI segment capital expenditure of $463 million in 2023, $5,633 million in 2024, $12,727 million in 2025, and $7,723 million for the three months ended March 31, 2026 against $2,567 million in the prior-year quarter. Space segment $1,497 / $2,032 / $3,832 million and Connectivity $2,455 / $3,498 / $4,178 million for 2023 / 2024 / 2025. These sum to $4.4 billion, $11.2 billion, and $20.7 billion, which reconcile exactly to the three capital-expenditure bars carried in the filed roadshow presentation. The prospectus attributes the increase, in each period and in identical words, to “the rapid expansion of our terrestrial data centers, including the development, construction, and equipping of new facilities and supporting infrastructure.” That language attributes the increase; it does not state that the segment contains nothing orbital, and no such claim is made here.

(14) Street targets. REPORTED, and none of it FILED: bank targets reach the record through coverage, and the notes themselves are not public. Morgan Stanley (Adam Jonas), $300 base target with a $600 bull case. JPMorgan initiated Overweight at $225. HSBC initiated Hold at $115 on July 24, 2026, with a blue-sky case of $293 and an estimate of about $106 billion of cumulative cash burn before free cash flow turns positive around 2030; reported through coverage (TheStreet, Yahoo Finance, MarketScreener, CNBC, July 24, 2026), with the note itself not public and the analyst not named in the coverage reviewed. All reported through coverage; the banks’ per-segment breakdowns are not public.

(15) The businesses and the contracts. FILED. SpaceX Form S-1 and final prospectus (accessions 0001628280-26-036936 and 0001628280-26-042639). The Anthropic compute agreement (about $1.25 billion a month) and the Google agreement (about $920 million a month, via free writing prospectus), both cancellable on ninety days’ notice, are disclosed there and in related free writing prospectuses; the SEC staff cited Item 601(b)(10) on the Anthropic agreement (comment-letter correspondence, accession 0000000000-26-005505). The pace-of-spending language is the company’s own: “During this investment period, our capital expenditures will scale as quickly as we are able to deploy power and compute to address the $26.5 trillion potential market opportunity for AI,” and “we believe speed is a competitive advantage... we continue to prioritize execution speed, capacity expansion, and technological leadership over near-term margin optimization.” The just-in-time characterization is the chief financial officer’s, in a separately filed interview.

(16) Governance. FILED (S-1; read in The Three Layer Cake: SpaceX’s Governance Structure): a controlled company with two share classes, one vote a share and ten; the founder controlling about 85 percent of the vote on about 42 percent of the economics, the public about 15 percent of the vote on about 60 percent of the economics; exempt from the independent-board-majority requirement; the controller elects and removes the controlling directors; and a shareholder proposal requires 67 percent of the voting power.

(17) Cursor. FILED. Acquisition disclosed before the offering, settled in stock (Form 8-K, accession 0001628280-26-043411).

(18) The bond. FILED for the terms, REPORTED for the trading. Debut $25 billion offering, five tranches, priced June 22-23 and settled June 26, 2026 (Form 8-K, accession 0001628280-26-044955): $7.0 billion at 5.35% (2031), $6.0 billion at 5.65% (2033), $6.0 billion at 5.875% (2036), $2.5 billion at 6.60% (2046), $3.5 billion at 6.65% (2056); weighted average coupon about 5.855 percent, about $1.46 billion of annual interest, semiannual on January 15 and July 15, first payment January 15, 2027. The offering refinanced and extended the March 2026 bridge loan (about $20 billion at roughly 4.6 percent, maturing September 2027), a debt-for-debt transaction rather than new deployable cash; separately, the company committed in its filings to a $25 billion minimum cash balance, which the agencies credited as supporting the rating, immobilizing that capital (read in Cursor Stock and the Investment-Grade Refinancing). Reported paper losses of about $305 million within days of pricing; the 2036 tranche’s yield toward 6 percent and spread above 160 basis points; the “junk-like despite investment grade” characterization (Connect Money, The Motley Fool, Advisor Perspectives, June-July 2026). Investment-grade ratings assigned June 18, 2026 (Moody’s Baa1, Fitch BBB+, S&P BBB).

(19) Operating figures. FILED. First-quarter 2026 net loss of $4.28 billion on $4.69 billion of revenue; capital expenditure of about $10.1 billion in the quarter, of which about $7.7 billion was AI. The funding gap and the compute-contract terms are read in earlier entries in this series: The $235 Billion Cash Gap, The Ninety-Day Annuity, and The Probability of Failure.

Analysis: Cape Fear Advisors.

Our $235 billion cash gap estimate has travelled far, a number others have reached for, and used to mark the distance between a filing’s prose and its math, between what a document discloses and what it does not. It was read from the uses side: the commitments the S-1 disclosed through 2030, each correct in its own note, larger in sum than the offering could raise. The sources and uses behind it have moved since, the proceeds among them, and this note does not re-price the whole bridge. It takes two lines we separated then and can read on their own, apart from Terafab, which we track on its own calendar: the AI-segment revenue, the compute the company agreed to sell, and the AI-segment capex, the build that revenue takes. We read these two contracts once before, for their timing and their weight in the price the offering asked. The uses side was read then. This is the revenue side.

The S-1 warned in prose and quantified where it disclosed the two compute contracts. Capital needs are significant, funding depends on the markets, cash flow will stay negative: warnings without figures. The two contracts carried the figures. Anthropic and Google, disclosed in detail, their headline values on the page and into the reporting on the offering. The company made these commitments specific, specific enough to carry into that reporting, and that specificity is what a later filing gets measured against.

The headline is the ceiling

The headline is the most the arrangement can pay. Anthropic runs $1.25 billion a month, about $15 billion a year, a monthly fee the filing carries through May 2029 and repeats in more than one place. Google, $920 million a month. Carried to their terms, the two headline near $45 billion and $30 billion.

Then the conditions take it back, set out in the terms below the headline. Both contracts cancel on ninety days’ notice, Anthropic’s from signing, Google’s after December 31, 2026. Both ramp at a reduced fee, Anthropic in May and June, Google through September, and in each case the reduced fee carries no amount. The full monthly number arrives only when both the fee and the delivered capacity reach full, and the capacity depends on a build not finished until the fall, so the full rate lands months after the fee ramp ends. The months before run lighter on both counts, by amounts the filing does not state. Priced against the terms, the most either side can be held to is about $3.8 billion for Anthropic, its first ninety-day window, and about $5.5 billion for Google, whose cancellation cannot start until after December 31, leaving a committed window nearer six months. Both fall further once the unpriced ramp is honored. A ceiling on the page, a fraction of it committed, and the difference left blank. Headline forward, the reductions in the terms behind it.

The two readings sit in two rooms. The S-1 carries a monthly fee through May 2029, which reads as three years. On May 28, on X, the chief executive called it “a 180 day lease with 90 day notice mutual cancellation thereafter,” said “the short term was our request, not Anthropic’s,” and added that “if compute gets super tight I said we might need it back.” Neither company filed a clarification. One room holds the annuity carried into the offering; the other holds a six-month lease the seller says it may take back. The filing runs long, the officer runs short, and the record has not squared them.

Material to sell, material to unwind

A filing cannot make these contracts central enough to sell an offering and ordinary enough to change without notice. The SEC staff drew that line: reviewing the registration, on the Anthropic agreement it asked what consideration the company gave to filing the agreement, and cited Item 601(b)(10). That rule exempts an ordinary-course contract from being filed as an exhibit unless the business is substantially dependent on it. A single contract worth about $15 billion a year, set against $18.7 billion of total revenue in 2025, is a contract the business leans on, whatever label the filing gives it. Material by the staff’s citation and by the company’s own size, and a material contract’s expected delivery does not change quietly. Material to sell, material to unwind.

The nearest date is capacity

Delivery comes before cancellation as the first test. The plant provided about a gigawatt of compute (filed), reached in the first quarter of 2026. The Google agreement, whose free writing prospectus sets its capacity at about 110,000 GPUs, runs a reduced fee through September and starts the full $920 million a month in October, behind a GPU delivery deadline of September 30. The timing only fits capacity that is not yet built. A seller with 110,000 GPUs already running would bill the full rate at once, not wait for October, and the delay leaves about $2.76 billion, three months at $920 million, of full-rate billing on the table. No seller forgoes $2.76 billion to postpone the meter on capacity already running. The October start is the tell: the capacity earns full rate only once it exists. The remedy corroborates it, since the agreement lets Google terminate or take fewer chips at a reduced fee if the September delivery is missed, a failure case written for hardware not yet in place. The existing gigawatt was not sitting free in any case: the Anthropic agreement already spans both clusters, and the company’s own risk factors say it needs access to significantly more AI chips than are currently available to it. The revenue waits on the build, and the build comes first. The reduced-fee ramp reads the same on both contracts, on overlapping clocks: the filing carries the ramp, not the reason.

That forks by September 30, harder and sooner than ninety days. To make the Google date SpaceX builds the capacity, capital out now against the gap; or it misses, and after a month’s grace Google may terminate or take fewer chips at a reduced fee, either of which discloses; or it serves Google out of what the Anthropic lease already covers, and that changes what Anthropic bought, which discloses too. Build, miss, or move it, and two of the three are material changes to a material contract’s delivery.

The bridge builds itself out of the job

The generous case: SpaceX spends what it takes and delivers both, in full, on time. It has then built for demand with an end date on it. The contracts run to May and June 2029 at the outside; the cancellation and the officer’s own six months pull the real line nearer; and the plant outlasts the revenue. That is the operator’s problem this series has read before, a long asset against short, cancellable demand, on SpaceX’s book this time.

And the customer is standing up its own supply. On July 22, the day this piece is dated, Anthropic and AMD announced up to two gigawatts of AMD capacity for Anthropic, the first gigawatt in the first half of 2027, with an AMD equity stake in Anthropic of up to $5 billion (reported). TeraWulf’s roughly 401 megawatts land for Anthropic late in 2027 (filed). On marginal-cost dispatch, capacity a buyer owns or has already paid for would run ahead of a rental at $1.25 billion a month, which puts Colossus at the margin, the first shed when the owned capacity comes on. That capacity arrives across 2027, so this year is captive and the displacement is a 2027 question. SpaceX built the bridge; the far bank is going up under the customer’s own hand. The build case pays, now, to stand a bridge meant to be idle the day the traffic it carries has better ground to stand on.

Where it resolves

We do not know which way SpaceX goes, and we are not calling it. These are the outs we can see; there may be others we cannot, and an out we did not name would be its own tell, because on this record every resolution files. Two filings are close. The mid-August 10-Q is the one that must: its capex will show whether the build was made, a large number the ordinary-course case and the gap feeding, a small one a silence on a headlined obligation; and ASC 280 requires disclosure of a customer past a tenth of revenue, which Anthropic clears several times over, so the quarter against the deal comes through whether or not the name is printed. How much comes through is a separate question. The Anthropic months in the quarter ran at the reduced ramp fee, so the specifics of consumption and price may be advanced or may fall beneath the threshold for a first partial quarter; the relationship is material, its opening months need not be. The concentration shows; the precision is the filing’s call. Over both sits the Item 303 duty on known trends, now carrying the day’s AMD fact. An interim 8-K is the open window: a termination is an Item 1.02 event and a material amendment an Item 1.01 event, prompt and required either way.

As of this writing, July 22, no such notice is on the record, and no 8-K has been filed, though an 8-K runs up to four business days behind its trigger, so the record speaks with a short lag. On our reading, cancellation runs ninety days from signing, so a change meant to reach the October step to full rate would have gone out around July 1, and none has. On the CEO’s reading, a 180-day base, the notice window has not even opened, so October is firm either way. The working read is that the plan holds: the company expects to deliver the full $1.25 billion a month to Anthropic around October, once the build completes, on top of what Google is owed and what the company’s own models take. That is the load the September 30 build has to carry, and the capex line and the concentration disclosure are where carrying it, or failing to, will show.

The positioning is where the choice shows. A change might be announced as internal need reclaiming the capacity, the line the officer laid down on X in May, or as a new relationship that forced the old ones to give. The story does not change the mechanics. Any change to either commitment takes the ninety-day notice the contracts name, and a notice that materially changes a material contract discloses. The story can set how the change reads, not whether it files. Intent posts to X; the commitment changes only on the record.

So the question stays open, and the answer is legible however it lands. The offering carried the ceiling, the contract commits ninety days, the customer is standing up its own supply, and squaring those three is a filed act every time. The record has two dates to speak, and how a company that made a large matter of these commitments settles them is the question the record answers.

Standing disclosure: Anthropic is the developer of Claude, which is used in preparing this research, and Anthropic is a counterparty to the two contracts read here. That nearness cannot be fully checked away, which is why no claim in this piece rests on trust in the tool: every figure carries a public source, and the record grades the rest. Others named in this piece have ties to Anthropic: Google is the second counterparty, and its parent Alphabet holds a large position in Anthropic; AMD, whose capacity announcement is noted here, is taking an equity stake in Anthropic; and TeraWulf leases capacity to Anthropic. Companies not named here, among them the GPU and cloud suppliers behind the capacity, may hold positions or supply relationships that bear on the filers discussed, and that possibility is part of why every piece is re-checked for bias, ground facts, and filings rather than read against a fixed list. Figures are quoted from the filers without characterization, and the same standard of reading is applied to every filer named.

NOTES

• SpaceX Form S-1, accession 0001628280-26-036936; final 424B4, accession 0001628280-26-042639. FILED. Prospectus Summary: 2025 revenue “on a consolidated basis of $18,674 million” (about $18.7 billion), and the clusters providing “approximately 1.0 gigawatt of compute power.” The Anthropic Cloud Services Agreement at $1.25 billion a month is disclosed in the S-1 and quoted verbatim in the comment letter below; its “through May 2029” term is carried in the offering free writing prospectus (0001628280-26-040610). The company’s risk factors state it needs access to significantly more AI chips than are currently available to it, the shortfall Terafab is meant to address. Pinned to sections and documents, not to prospectus page numbers, which the EDGAR HTML does not carry.

• SEC comment-letter correspondence from the Division of Corporation Finance, UPLOAD accession 0000000000-26-005505 (May 29, 2026), verbatim: “We note the revisions made in response to prior comment 3, including that Anthropic will pay you $1.25 billion per month for access to compute capacity across COLOSSUS AND COLOSSUS II. Please tell us what consideration you gave to filing this agreement. See Item 601(b)(10).” FILED.

• Google Cloud Services Agreement: disclosed via a Rule 433 free writing prospectus, accession 0001628280-26-041150, filed June 5, 2026. It states the agreement was entered June 5, 2026 and provides for $920 million per month from October 2026 through June 2029 across about 110,000 GPUs. The agreement conditions Google’s obligation on delivery of the committed GPUs by September 30, 2026; on a missed delivery, after a one-month grace period, Google may terminate or take fewer chips at a pro-rata-reduced fee (per “What the Compute Contracts Commit”). This sources the exhibit’s Google term (”through June 2029, FWP language”) and is consistent with the FWP fact carried in “What the Compute Contracts Commit.” The Anthropic monthly fee and the Terafab framework also appear in the June 4, 2026 roadshow free writing prospectus (accession 0001628280-26-040610) and the UK retail free writing prospectus (accession 0001628280-26-040874); Terafab is set aside in this piece and read on its own calendar.

• Elon Musk, X, May 28, 2026: the 180-day characterization and the “might need it back” statement. REPORTED.

• Anthropic and AMD, up to 2 GW of MI450-series capacity, first gigawatt H1 2027, AMD equity investment up to $5 billion in Anthropic (AMD investor relations, July 22, 2026, ir.amd.com/news-events/press-releases/detail/1292/; no SpaceX 8-K filed as of that date). REPORTED.

• TeraWulf Form 8-K, accession 0001104659-26-080583 (the ~401 MW, twenty-year Anthropic lease, delivery late 2027 to 2028). FILED.

• “What the Compute Contracts Commit” (June 6, 2026), this series, which carries the contract terms, the ramp months, and the May 3 and June 5, 2026 signing dates shown on the exhibit; and “The $235 Billion Cash Gap” (May 21, 2026).

• ASC 280-10-50-42 (customer-concentration disclosure); Item 303 of Regulation S-K (MD&A trends and uncertainties); Form 8-K Item 1.02 (termination of a material definitive agreement) and Item 1.01 (entry into or material amendment of a material definitive agreement).

Figure unchanged from “The $235 Billion Cash Gap“ (May 21, 2026). Circled: the Terafab estimate, and the disclosure that carried it. The estimate stands; the disclosure now has an executed agreement, eight state applications, and a prospectus record behind it.

Four forms of commitment, one signature

Announced: up to $119 billion. The county’s posted hearing notice carries the number in the county’s voice: $55 billion for the initial phases, $119 billion total if all phases are built. The number has a dated history. The idea surfaced at Tesla’s November 2025 shareholder meeting. The March announcement placed a roughly $20 billion fab near Austin, a vintage preserved in the record because T1 Energy, a solar company, filed a March investor deck citing “a $20 billion chip manufacturing Terafab in Austin” as a demand tailwind. April’s coverage of Intel joining carried $25 billion. The May application to Grimes County reached $55 billion initial and $119 billion total. The announced number roughly sextupled in eight weeks, each step dated in a filing or a release. (1)

Framework: a characterization, without a filed text. The S-1 describes what stands behind the announcement: a “general framework” with Tesla under which specific projects, “including any development timelines, milestones and capital expenditures,” remain “subject to separate negotiations and agreements” and “have not yet been determined.” The risk factors go further: “neither Tesla nor Intel are obligated to remain a part of the project, and we may not enter into any such definitive agreements.” The exhibit index completes the picture. The S-1’s filed material contracts run from the indemnification form through the compensation plans, the Musk restricted stock agreements, the EchoStar spectrum purchase, the bridge loan, and the xAI merger. Item 601 requires material definitive agreements as exhibits, so the framework’s absence from the list is the issuer’s own judgment on what it is: prose. (2)

Application: $55 to $119 billion, sworn by a subsidiary. The applicant in the eight state filings is TeraFab AI, LLC, Texas taxpayer 32105982659, at 1 Rocket Road, Starbase, Texas, parent company SpaceX alone. Its project description calls the work “led by a consortium of affiliated advanced technology companies, including Tesla, Inc., Space Exploration Technologies Corp. (SpaceX), and xAI Corp.” Intel, a public member of the consortium since April, is absent from that list; “affiliated” is doing precise work. The eight applications, four phases across two school districts, seek roughly $1.66 billion of school maintenance-and-operations tax benefit through 2046. Each commits to ten jobs, the statutory floor under the JETI framework, at a wage floor of $66,668: eighty jobs across all eight filings, beside the county headline of 1,800. Both are filed commitments to government bodies, weeks apart. At every layer of the record, the binding number is a small fraction of the announced one, for jobs as for money. (3)

Executed: the county agreements, and the price of leaving. The 312 tax abatement agreement and its companion 381 economic development agreement took effect June 3; SpaceX’s chief financial officer signed June 22, per the records KBTX obtained and posted. The binding terms, from the agreement text: invest at least $5 billion in facilities and equipment by the end of 2030, with a 90 percent safe harbor that makes $4.5 billion the real floor; create 1,800 full-time-equivalent jobs by the end of 2035, transfers from other Musk-company facilities expressly counting; pay the county $20 million a year in lieu of taxes for 2027 through 2036; and after 2036, under the 381, the county collects full taxes and grants back everything above $20 million a year through 2061. One payment is unconditional: $10 million, due by August 2. (4)

The exit clause, Section 10.13: SpaceX may terminate “at any time and for any reason” on 30 days’ written notice, and termination “shall not constitute a default or breach... nor shall it give rise to any claims for damages, penalties, or recapture,” except that the company repays abated taxes for up to three years, net of PILOTs paid. Early reporting rendered that as a $60 million cap, three years of the $20 million payments. The clause keys the repayment to abated taxes, a function of what sits on the tax roll. Today the roll is empty, so the exit price today is the $10 million payment, and the agreement waives the 2027 PILOT for a termination before June 1, 2027. The same clause grows with the build: three years of abated taxes on a completed campus would run to hundreds of millions. The walk-away costs least during the years the project is a story, and starts costing money on a date the agreement names.

What four public companies tell the SEC about a $119 billion project

The four federal records, each in its filer’s own materiality judgment. SpaceX’s final prospectus mentions Terafab 51 times, in the narrative and the risk factors, and once in the audited financial statements, a single sentence in the related-party note recording that the company “announced a collaboration with Tesla to build a chip manufacturing facility (referred to as Terafab).” Tesla, the framework counterparty, has yet to use the word Terafab in any SEC filing: the 10-K, the 10-K/A, the first-quarter 10-Q, and two earnings exhibits, all filed since the March announcement, discuss SpaceX itself while leaving the project unnamed. Intel’s entire federal record on the project is one line in a furnished earnings release from April 23: it “joined the Terafab project as a strategic partner alongside SpaceX, xAI, and Tesla,” a membership list that matches the state application’s Musk trio with itself appended. T1 Energy carries the stale March vintage as a market tailwind. (5)

Is Intel in or out? By venue. Intel appears in the April announcement, in the technology story, the controlling shareholder saying the fab will run Intel’s 14A process, in the personnel flow, and in the county’s correspondence file. Every binding document tells the other story: the sworn consortium omits it, the audited related-party sentence names Tesla alone, the risk factor releases it from any obligation to remain, and its own filings give the project one furnished line.

The omission works backward. “Affiliated,” in a sworn application, is a word someone applied, and applying it correctly sorted the consortium: as sworn to the State of Texas, the project is led entirely by companies under one man’s control. Tesla holds the strangest seat in that sorting, the most included member in the narrative, named as family by a subsidiary in a venue where it bears nothing, and the least bound in the documents, since the county agreement’s own affiliate definition, a majority-control test, would leave it outside. On June 30 the project acquired its first named leader, Gary Jiang, a seventeen-year Intel manufacturing veteran who ran tool installation and ramp for Intel’s 18A node, hired as “Director, Terafab.” Tesla hired him. (6) The fab’s first named director draws a paycheck from the consortium member whose federal filings, through July 7, have yet to use the project’s name. Tesla reports its second quarter in late July; the silence has a test date.

The question matters because of what is missing. The record shows most of the stack assembled, and assembled inside the controller’s reach: land in the deeds, water in the rights, power in the county agreement’s own gas plants, the tax structure executed at the county and certified at the state, an entity, internal demand, and chip design teams inside the family. Three pieces remain, and each has one source. The tools: ASML alone makes the extreme-ultraviolet lithography machines a leading-edge fab requires, and the courtship is on the record, the controlling shareholder addressing ASML’s employees on Terafab within days of the Dimon conversation. (7) The process: TSMC and Samsung guard their nodes, Intel licenses, and the reported deal runs on Intel’s 14A. The operating knowledge: a licensed node is ramped by people trained on it, that experience lives on Intel’s payroll, and it has begun to move, one director so far. TSMC offers the alternative to building at all, wafers bought rather than fabs owned, and the company’s own free writing prospectus forecloses it: even the best-case assumptions of existing manufacturers fall short. So an owned Terafab requires Intel’s process and Intel’s people, whatever TSMC sells, and two of the three single sources trace to the one consortium member every binding document leaves free to go. The forward watch writes itself: a process license would be the project’s first definitive agreement anywhere, and it would reach the federal record through Intel’s filings, from the member the sworn record excludes.

The census: a project publicly sized at up to $119 billion whose total binding federal disclosure is one risk factor warning the definitive documents may never exist, one sentence in an audited footnote, one furnished bullet, and silence from the framework counterparty. The local record extends the posture: on June 25 the county asked the Texas Attorney General for permission to withhold its SpaceX, Tesla, and Intel communications from KBTX’s records requests, and confirmed along the way that the abatement rests on no independent economic or feasibility analysis. The state record has the timing model and the entity. The county record has the only signature. The federal record has the disclaimer.

The applicant’s own schedule: the spending lands in 2028 through 2037

The state applications carry what the federal record lacks: a year-by-year model. Each application’s Attachment C is a 20-year schedule of the taxable value the project expects on the ground, filed with the Comptroller. These are taxable-value schedules, a proxy for deployment timing net of depreciation and assessment mechanics, and the label matters: this is the company’s own filed model of what property it expects on the ground, by year. (8)

By its own filed schedule, essentially nothing arrives in 2026: $40 million of property, then $176 million in 2027, $4.2 billion in 2028, and $10.7 billion in 2029 at Phase 1’s peak. Phase 2 peaks in 2032 at $21.8 billion, Phase 3 in 2035 at the same figure, Phase 4 in 2037 at $17.3 billion. Value on the ground crests near $45 billion in 2036 and declines as equipment depreciates, modeling to roughly $22 billion by 2045, against which the county’s take is $20 million flat.

August reads clean. The first 10-Q is due by mid-August under the 45-day rule, and the second-quarter capital expenditure line is the one number that has to satisfy two owners at once: the growth story needs it rising, the rating math needs it contained. The filed schedule settles what the line covers. Terafab contributes $40 million of 2026 property by its own model, so whatever the line prints, Colossus and the data centers are carrying it, and Terafab’s real cash calls begin in 2028.

The heavy years land inside the rating window. The agencies rated the buildout as something the company can choose to slow: S&P models negative free cash flow through roughly 2030, and Moody’s requires leverage below 3.25x by end-2028. (9) The filed schedule places Terafab’s heaviest deployment, 2028 through the early 2030s, inside that window. The company’s own state filing schedules its largest discretionary program’s peak spending into the years its rating covenants require the balance sheet to improve. Both sides are filed.

The covenant clears with room to spare. The county’s binding floor, $5 billion by 2030 with the safe harbor at $4.5 billion, sits at roughly a quarter of the $18 billion the company’s own model shows on the ground by then. Half-failure would still clear it.

One clause, two markets

The walk-away clause carries two prices at once, both on the record. The rating agencies credit SpaceX with discretionary spending it can defer, and Section 10.13 is that discretion in contract form; the clause supports the floor and the grade. The equity prices the hyperscaler story, which requires the build to be certain; the clause is the term the valuation assumes stays unexercised. Credit prices the option to stop. Equity prices the certainty of going. One document, two readings.

The models price the revenue and assume the plant

The revenue is in every model. Goldman, the lead underwriter, carries roughly $474 billion of total revenue by 2030, about $322 billion of it AI; Morgan Stanley, a co-lead, roughly $330 billion and $190 billion. The most generous credible independent, Aswath Damodaran, took the company’s story at close to full strength, doubling his AI revenue target to $160 billion by the mid-2030s and holding the probability of failure at zero, a setting his own pre-IPO practice reserved for companies that fund themselves, and reached $1.25 to $1.35 trillion, with the offer priced at 138 percent of his value. (10) The published bear, Morningstar, sits at $780 billion. Every build is a claim about AI revenue arriving on a schedule.

All of it is compute revenue. The filings route the fab’s output inward: chips “optimized for AI robots and vehicles” and chips “optimized for the space environment to be used in our orbital compute infrastructure,” inputs to the compute the AI segment sells and to products a different public company sells. What connects the fab to the revenue is the issuer’s own risk factor, one continuous passage: the company’s orbital AI ambitions “depend on our ability to access a sufficient number of AI chips, significantly more than are currently available to us”; “we expect to construct Terafab to address such supply constraints”; and should Terafab fall short, “we may not have other sources of sufficient AI chips.” Revenue, to chips, to the fab, to no alternative, in the company’s own sequence. The controlling shareholder’s filed conversation with Jamie Dimon generalizes it: the nearest American memory fab reaches volume in 2028, “even the best-case assumptions of existing manufacturers fall short of anticipated demand,” and “that’s why we need to do Terafab. It seems essential. Otherwise... there will not be enough chips.” One threshold: today’s contracted revenue rides third-party silicon; the compute SpaceX leases to Google under the June cloud agreement runs on 110,000 NVIDIA GPUs. (11) The fab becomes load-bearing at the ramp, the growth from today’s headline contracts to the models’ hundreds of billions, which requires compute at a scale the filings say only the unbuilt plant can supply. And the ramp keeps being sworn to new regulators: in January the company filed FCC plans for a million-satellite orbital data center constellation, and this week it asked the FCC to authorize 100,000 third-generation satellites as infrastructure for AI. (12) The demand side now sits in county, state, securities, and spectrum filings. The supply side remains a framework, and each new application enlarges the shortage the framework exists to cure.

Each revenue path therefore implies a capacity date, and the sworn schedule answers it. Revenue collected in 2030 requires compute in service around 2029 and chips before that; at the end of 2029 the filed schedule has one phase standing, $10.7 billion of property, with the volume phases arriving 2032 through 2037. The bull case needs the fab’s output roughly three years ahead of the applicant’s own schedule. The generous independent’s mid-2030s vintage fits the calendar, and the money sits outside his model by design: he reinvests about $133 billion across ten years, roughly $13 billion a year, funding the revenue he forecasts rather than the buildout the company announced, while the county filing prices this one project at $55 to $119 billion and the company’s actual capital spending ran $10.1 billion in the first quarter.

Damodaran’s spreadsheet carries four revenue lines, launch, Starlink, AI, and an expansion bucket he describes as “expansion options embedded in each business,” with Terafab appearing nowhere in the spreadsheet or the post. Assign the fab to that bucket as generously as possible and the assumptions require it to be the most capital-efficient fab ever operated, by a factor of about five. The bucket builds $100 billion of 2036 revenue with roughly $20 billion of capital and implies an after-tax return on incremental capital near 120 percent. TSMC, the most efficient fab operator in history, at the peak of its strongest cycle, turns roughly a dollar of invested capital into a dollar of annual revenue, spent 33 percent of its 2025 revenue on capital to hold that position, and prints returns in the high forties at the cycle top and the twenties through it. (13) The bucket’s modeled margin sits below TSMC’s actual one; the entire excess is the capital assumption. The bank models stay private, their reported summaries carry compute revenue and, in Goldman’s case, name “future equity raises” without quantifying them. The cash for the plant enters the record as an assumption in the one public model, a phrase in the lead underwriter’s, and silence in the rest. No published build prices the plant its own revenue requires. And the exclusions are sound practice, which is the point: an unsigned project is not a valuation input. The $119 billion lives in one place only, the price.

The cash: $80 billion deployable, against a dated $80 billion

Cash is the reconciling variable, because the equity rests on the revenue while the debt restrictions cap the spend: Moody’s leverage test at end-2028, S&P’s tolerance conditioned on continued raising, Fitch’s expectation that the company defers discretionary capital if access tightens. Debt at fab scale would spend the grade the company bought in June.

Capital coming in has not been the constraint. In its first three weeks public the company raised $86 billion of equity, sold $25 billion of bonds on $89 billion of orders, and agreed to buy Cursor for $60 billion in stock: capital in three currencies, roughly $171 billion of it. The June 22 bond disclosure put cash at $100.8 billion as of June 19; the bond settled June 26 and the bridge was repaid from it, leaving roughly $105 billion. The $25 billion minimum-cash commitment, the price of the rating, holds the floor, and roughly $80 billion sits above it. Against it, the dated requirements: the fab’s initial phases, the county’s own $55 billion definition, spent by roughly 2029 if the 2030 revenue is to exist; AI-segment capital spending already running $7.7 billion a quarter; the Spectrum close taking $11.5 billion of cash in November 2027; Valor leases near $3.4 billion a year; a free-cash-flow run rate near negative $36 billion a year. (14) Any conservative assembly of the project pieces alone puts at least $80 billion out the door before the revenue’s date. The deployable pool and the dated project requirement are the same size, before a quarter of operating burn touches either.

The other sources stay open, and each carries a price the company has already been paid once. The debt door has opened once already: the June bond, rated into the grade, retired the bridge and added about $5 billion of new cash; debt at fab scale beyond it spends the grade, with Fitch’s rating conditioned on deferring discretionary capital and the covenant counting cash against the leverage test. The $25 billion floor is spendable by un-committing it, and the agencies credited the commitment. The customers could prepay or commit, by reopening contracts whose value to the story is that they read as firm at the headline; the two anchors pay monthly, in arrears, on 90-day terms, and financed none of the construction. The partners could fund, by signing the definitive agreement that starts the disclosure machinery at both companies; the framework leaves capital expenditures “not yet been determined” precisely as long as it stays a framework. The county and state incentives run the other way in the near term, $10 million out on August 2. Every door reopens by handing back what it already bought: the grade, the floor, the headline, the silence. Equity is the door that costs only new money, and the S-1 reserved it in a single sentence: “We may issue a significant amount of equity in connection with future transactions.” The sentence sits inside the acquisitions risk factor, the paragraph that names xAI, the spectrum, Terafab, and Cursor, so the disclosed funding plan is equity-for-transactions, and Cursor proved the mechanism six days after listing: $60 billion, in stock. It comes in two shapes, a secondary offering or a stock-funded acquisition, and an acquisition could close the missing pieces and the cash gap in a single document. Either shape spends the same currency, which requires the price, which requires the announced number alive, which the calendar tests. Its filed complication comes with it: new shares would price from a level the most generous independent marks 28 percent lower, into the same window the lockup releases insider supply.

Also in the record: water and power

The campus supplies its own inputs, and the record shows who owns them. The power, from the county agreement’s own text: the project is “expected to include semiconductor manufacturing facilities; natural gas fired power plants, and artificial intelligence facilities to be used in conjunction with the manufacturing process,” powered on site and off the grid. The power buildout is inside the same capital requirement as the fab.

The water has a filed magnitude and an open owner. TCEQ records show water-rights certificates tied to Gibbons Creek Reservoir were amended in August 2023, nearly three years before the project was named, authorizing permanent consumption up to 3.2 billion gallons a year, several times what a large semiconductor plant draws, in keeping with a campus that is a fab, an AI facility, and its own power plants. On May 27, one week before the county vote, WIT TECH LLC, a Wyoming entity, bought six parcels totaling 2,796.5 acres, including the Navasota River pump station and the one-acre parcel that is the river itself, with reporting placing contracts for more than 6,000 acres behind them. (15) The entity’s paper trail points two directions: its Wyoming filings carry the signature of Jared Birchall, who manages the Musk family office, and its deed records carry James Burnham, xAI’s general counsel, and xAI has been inside SpaceX since February. The public record leaves the water’s balance-sheet home open: the controller’s private holding, or already the public company’s. The county agreement draws the facility’s water from Gibbons Creek Reservoir and names no owner of the intake. A future disclosure settles it either way: a related-party note if the water is private, a subsidiary schedule if it is home. KBTX traced the parcels and the entity, part of a records effort running past twenty public-information requests; these documents are public because of that work.

The application’s compelling-factor section argues the limitation is decisive because otherwise “competing jurisdictions (e.g., Arizona) would present a more favorable after-tax return.” The consortium’s public plan contemplates an Arizona fab beside Intel’s campus regardless, and the Comptroller certified the factor on June 15.

The clock: the dates that test the project, and the financing behind it

If the company steps back from Terafab, when must it say so? The disclosure rules attach to decisions and to filed agreements, and the record holds neither. The framework was never filed; the county agreement, though signed, sits outside the S-1’s exhibit list, so by the issuer’s own materiality judgment even a termination notice arguably travels without a report. The risk factors, written before the offering, already cover the exit: the project “may not be successful,” the definitive agreements “may not” be entered. The structure that never required a commitment never requires an un-commitment. A decision would surface in a later report, if a decision is ever made; drift is quieter than that.

What replaces disclosure is a calendar of small public payments and other companies’ reporting dates, each a revealed preference. July 13 and 14, approximately: the school boards vote, the Governor’s approval still pending, and since property placed in service before the agreements execute loses eligibility, a serious builder signs quickly. Mid-to-late July: the counterparties report before the company does. ASML discloses bookings, and extreme-ultraviolet tools are ordered years ahead of wafers, so a schedule that finishes construction in 2028 has to leave orders in that backlog within quarters; TSMC takes questions days later; Intel, which put the project in its first-quarter release, chooses among updating the line, repeating it, and dropping it, all three legible; Tesla’s silence gets its test. August 2: the $10 million pays, maintaining a $119 billion story at a dollar per twelve thousand, or a Material Obligation defaults in a public record. Mid-August: the first 10-Q, with the fab absent from the capex line by its own model, so the line reads on Colossus alone; the lockup’s first trigger runs two trading days behind the release. Year-end: the sworn applications put construction commencement in 2026, and the Comptroller’s letters rescind if the information changes. February 1, 2027: half the 2027 PILOT, $10 million. May 1, 2027: the first annual compliance certificate to the county. June 1, 2027: the exit gate closes; the agreement waives the 2027 PILOT for terminations before this date, so every later day is the first that leaving costs real money. June 15, 2027: the state recommendations lapse if the school agreements sit unexecuted.

Across the same months run the other two clocks. The lockup releases begin at the earnings trigger and run through 180 days in December and 366 days next June. The equity raise, the cheapest door the record leaves, has to come before the spending it funds, and the bull case needs the spending by 2028; the raise window is this window. Three calendars, one year: the announced number has to stay alive to price the raise, the county’s payment dates test whether it is, and the insiders’ first sales run alongside both. The market’s visibility into a $119 billion project runs through whether a $10 million check clears in Anderson, Texas.

Later still: the definitive consortium agreements, if they are ever signed, and the related-party notes of future filings, where the water, the subsidiary, and the allocation among four balance sheets each acquire a document. Until then the $119 billion is a number with one signature under one twenty-fifth of it.

Whence the cash. Still open, now dated. And, in whose name.

Part of a continuing read of the SpaceX offering documents, taken one at a time and in a single currency. Clause citations are pinned to the agreement texts; figures are verified against the primary filings. Analysis: Cape Fear Advisors.

NOTES

(1) Grimes County public hearing notice, county-posted: “Estimated capital investment for the initial phases is $55 billion, with an estimated total capital investment (if additional phases are constructed) of $119 billion.” The March vintage: T1 Energy investor presentation, EX-99.2, March 31, 2026 (”a $20 billion chip manufacturing Terafab in Austin”). The April vintage, $25 billion, per contemporaneous coverage of Intel’s joining.

(2) SpaceX Form 424B4, June 12, 2026: the framework language and the risk factor quoted appear in the business section and risk factors. The exhibit index to the S-1 lists the material contracts named; the Tesla framework agreement is not among them. Item 601 of Regulation S-K governs required exhibits.

(3) JETI applications J0035 through J0038 (Anderson-Shiro CISD) and J0039 through J0042 (Iola ISD), TeraFab AI, LLC, applicant, with Texas Comptroller recommendation letters signed June 15, 2026. Applicant identity, parent, jobs commitments, and wage floors per the applications; the $1.66 billion is the sum of the eight “estimated M&O gross tax benefit” figures in the Comptroller’s Attachment A summaries. Each recommendation is rescinded if the application is modified or the information presented changes, and lapses unless the school agreements are executed within one year.

(4) Grimes County 312 Tax Abatement Agreement and Section 381.004 Economic Development Program and Agreement, each effective June 3, 2026, as released by the county and posted by KBTX; the June 22 signature of SpaceX’s chief financial officer per KBTX’s report of the fully executed version. Terms cited: Sections 3.4 and 3.6 (payments), 4.1 through 4.3 (investment, jobs, safe harbor), 10.13 (termination), and the 381’s Section 3(a) (grants of taxes above $20 million, 2037 through 2061).

(5) EDGAR full-text search, July 7, 2026, all issuers, “Terafab.” SpaceX: Form 424B4. Tesla: 10-K, 10-K/A, first-quarter 10-Q, and two 8-K exhibits filed since March 2026, each discussing SpaceX without naming the project. Intel: Q1 2026 earnings release, Exhibit 99.1, furnished April 23, 2026. The county’s Attorney General request and the absence of an independent feasibility analysis per KBTX, July 2, 2026.

(6) The director hire: Electrek and Tom’s Hardware, June 30, 2026, from the executive’s own announcement of the role.

(7) The ASML address: Bloomberg, June 6, 2026, and CNBC, June 11, 2026. The 14A process: the controlling shareholder’s public statements and contemporaneous reporting.

(8) Attachment C of each JETI application, aggregated across both school districts and all four phases. The schedules state taxable value by year; they are a timing proxy, net of depreciation and assessment mechanics, and are not capital expenditure. Iola Phase 2 figures mirror the Phase 3 schedule shifted by three years, consistent with the OCR of the filed tables.

(9) Moody’s, Fitch, and S&P rating actions, June 18, 2026, as described in the June 19 note in this series.

(10) Aswath Damodaran, post-prospectus SpaceX valuation, June 2026, posted with the full spreadsheet. The failure-probability comparison draws on his prior pre-IPO valuations (Uber, Lyft, WeWork, Airbnb, Zomato, at 5 to 20 percent) and his Tesla series, where zero arrived only with self-funding; the June 7 note in this series sets out the record.

(11) SpaceX free writing prospectus, June 5, 2026: the Google Cloud Service Agreement, approximately 110,000 NVIDIA GPUs, $920 million per month from October 2026, terminable by either party on 90 days’ notice after December 31, 2026. Terms read against the headline in the June 6 note in this series.

(12) FCC filings: the orbital data center constellation application, January 2026, accepted for filing by the Space Bureau in February; the third-generation constellation application, approximately 100,000 satellites, submitted the week of July 6, 2026, per the filing and contemporaneous coverage.

(13) TSMC 2025 results ($122.9 billion revenue; $40.9 billion capital expenditure) and 2026 capital guidance ($56 billion), company reports and press coverage; returns as computed by standard screens, cycle-peak versus through-cycle.

(14) Cash and capital, as filed: the Form 8-K launching the notes offering, June 22, 2026, disclosed approximately $100.8 billion of cash and cash equivalents as of June 19, 2026; the pricing 8-K of June 23 set the five tranches ($7.0 billion at 5.350% due 2031, $6.0 billion at 5.650% due 2033, $6.0 billion at 5.875% due 2036, $2.5 billion at 6.600% due 2046, $3.5 billion at 6.650% due 2056); the offering settled June 26, with proceeds repaying the bridge loan in full, paying fees, and the remainder to “general corporate purposes.” The Cursor consideration is stock (Form 8-K, June 16), capital in without cash. Requirement components as filed: first-quarter 2026 consolidated capital expenditure of $10.1 billion, $7.7 billion of it in the AI segment (424B4); the EchoStar spectrum cash component, $11.5 billion (Note 6); the Valor lease obligations (Note 17); 2025 free cash flow of negative $13.9 billion and the first-quarter run rate, per the cash-flow statements. The arithmetic follows the June 5 and June 19 notes in this series.

(15) TCEQ water-rights certificates and amendments, and Grimes County deed records, via KBTX’s reporting, with additional deed detail (the xAI general counsel’s appearance in the records; contracts for more than 6,000 acres) via national pickup of the same records.

On June 16, SpaceX filed a Form 8-K converting its April option over Anysphere, the company behind Cursor, into a binding merger agreement. The deal is signed, in stock: an implied equity value of $60 billion, paid in SpaceX Class A shares, with the exchange ratio set by the volume-weighted average price over the seven trading days before closing. The option had set the choice as $60 billion in stock or a $10 billion cash payment to walk away. The merger takes the stock, and closes in the third quarter.

The form is the point. The consideration is shares, valued at the close, so the number issued floats with the price: the higher the stock trades into closing, the fewer shares change hands. It is the first use of the new public equity as currency, and the prospectus had named it. The risk factor on integrating acquisitions reserved the right to issue a significant amount of equity in connection with future transactions, and six days after the shares began trading, that sentence was a $60 billion deal. The business it brings is, by its own reporting, fast-growing and not yet profitable, its revenue consumed by the cost of the compute it runs on. That compute is bought today from other AI companies; brought in-house, it adds to the build SpaceX must fund. Either way the cash question holds. The acquisition is settled in equity, and the cash requirement stands where it stood.

On June 18, all three agencies rated SpaceX investment grade: Moody’s at Baa1, Fitch at BBB+, S&P at BBB, each with a stable outlook, all at the lower end of the scale. The agencies were specific about the basis. Moody’s rated the company under its telecommunications methodology, because Starlink carries the majority of revenue and earnings, and named Starlink the principal source of cash. The rating, Moody’s wrote, is limited by the large-scale AI buildout: its high capital intensity, persistent negative free cash flow, and an unpredictable range of returns.

The other two spoke to the cash directly. S&P projected negative free cash flow through 2029 and wrote that the company will need to raise additional capital through debt and equity markets to cover the deficits, the IPO proceeds financing part of the shortfall. Fitch put a condition on the rating: it expects the company to defer discretionary capital deployment if access to capital were limited. All three described the negative free cash flow as elective growth investment rather than a structural weakness. They rated the buildout as something the company can choose to slow.

That is the rating in the agencies’ own terms. It rests on Starlink as the cash engine, names the buildout as the constraint, and expects the company to keep raising and to slow the build if it cannot. The grade affirms the cash requirement rather than meeting it. It is given on the expectation that the requirement keeps being met, and on the company’s room to step back from the build if it is not.

The rating came at a price the company set itself, and it is paid in cash. In its filings SpaceX committed to hold a minimum cash balance of $25 billion, and the agencies credited that commitment as supporting the rating; the bridge loan’s covenant already counts unrestricted cash against the leverage limit. So $25 billion of the company’s cash is now spoken for, committed to the balance sheet to hold the grade, and held back from the buildout. The rating that lowers the cost of borrowing is the rating that freezes the cash. Following the money, that is the week’s largest movement: not cash raised, but cash immobilized.

On the same day, SpaceX’s bankers began arranging the company’s first investment-grade dollar bond, expected the following week, at a size reported as at least $20 billion and not yet fixed. The bond is required. The bridge loan SpaceX took in March, after acquiring xAI, matures September 2, 2027, and the prospectus obliges the company to apply IPO proceeds to repay it within six months of the offering. The bond refinances that bridge, new debt for old, and settles a maturity already on the clock. It leaves the cash requirement where it stood. Should the bond price above the amount the bridge needs, the excess would add to available cash, at a spread that is next week’s figure.

So the week set down three documents. The acquisition is signed, in stock. The ratings are investment grade, resting on Starlink, naming the build as the constraint and the raising as the expectation, and carrying the $25 billion the company has committed to keep idle. The bond refinances a bridge already due. The two figures the week put in play are circled above: the $60 billion settled in stock, and the $20 billion refinanced in debt. Both went to other ends; the buildout drew on neither.

Whence the cash. Still open, and now with fewer ways to answer.

— — —

Part of a continuing read of the SpaceX offering documents, taken one at a time and in a single currency. Sources: SpaceX Form 8-K, June 16, 2026 (Anysphere merger agreement); Moody’s, Fitch, and S&P rating actions, June 18, 2026; SpaceX Form S-1 and the SpaceX Bridge Loan credit agreement; press reporting on the bond offering. Figures verified against the primary filings. Analysis: Cape Fear Advisors.

The mechanics ran on schedule. The registration statement was declared effective on the morning of June 11. The company priced at a fixed $135 per share — no range, no book-built discovery — and filed its final prospectus the morning of the trade. On June 12, SpaceX began trading on Nasdaq under SPCX, opening sharply above the offering price and taking its place among the largest public companies in the world by market value. With the underwriters’ option exercised, the offering raised approximately $86 billion. By every market measure, the largest public offering in history succeeded.

On June 8, in a recorded conversation with Jamie Dimon filed with the SEC as a free-writing prospectus, Elon Musk was asked what compelled him to build chip fabs now, alongside everything else he was doing. His answer placed Terafab at the center. He described the limiting factor as the ability to make chips — logic, memory, and packaging — and said there is not a single high-volume computer memory fab operating in America, that the nearest, Micron’s in Idaho, will not reach volume production until 2028, and that even the best-case assumptions of existing manufacturers fall short of anticipated demand. His conclusion: “that’s why we need to do Terra Fab. It seems essential. Otherwise... there will not be enough chips.” In the registration statement, Terafab had been a collaboration the company was under no obligation to complete. Three days before the offering, the company’s controlling shareholder called it essential.

Figure unchanged from “The $235 Billion Cash Gap” (May 21, 2026). In the filings, the Terafab line was a collaboration with no commitment. On June 8, the company described it as essential.

On funding, the same filings stayed at the level of intention. The company wrote that it “plan[s] to access a range of debt and equity financing solutions available to us as a public company to fund future investments in growth,” and that its capital expenditures “will scale as quickly as we are able to deploy power and compute.” The requirement itself was never quantified. The tax-abatement filing the company submitted to Grimes County in May had placed Terafab’s first phase at $55 billion — the figure carried in the table above — while the offering documents named no Terafab commitment at all. And in the same disclosure that carried the necessity, the company set down its conditions: “while we expect to construct Terafab to address such supply constraints, Terafab may not be successful,” it “may not be able to achieve our objectives with respect to Terafab within the expected timeframes, or at all,” and “neither Tesla nor Intel are obligated to remain a part of the project, and we may not enter into any such definitive agreements.”

That is where the week leaves it. The opportunity to raise cash came and went, with a clear success: the offering priced, traded, and closed among the world’s most valuable companies. In the same days, the company moved Terafab from a collaboration it was not obligated to complete to a project its controlling shareholder called essential — while the documents that would fund it named no commitment, quantified no requirement, and conceded the project may not proceed at all. The need is arriving sooner than it was a week ago. The sources are no nearer. The question that opened the week is the question that closes it: whence the cash?

The input

In Professor Damodaran’s framework, the probability of failure is a standard input: the chance, over the foreseeable future, that the company does not survive as a going concern, applied as a haircut to the computed value. The model leaves it switched off unless the user turns it on. In his SpaceX valuation, in April and again in June, it is off, at the model’s default of zero. In the same block of inputs he turned on the two assumptions directly above it, the cost of capital and the return on capital after year ten, and entered bespoke values. The failure input stayed at the model’s default, which states the assumption in the model’s own words: no chance of failure over the foreseeable future.

What earns a zero

Professor Damodaran has switched that input on before, and moved it, for a single company over time. He first valued Tesla in 2013 and has valued it almost every year since, and the failure probability tracked the company’s finances.

In June 2019, with a debt overhang he called “a clear and present danger” and a real chance of the company losing access to new capital, the probability of failure was set to 20 percent. As Tesla steadied it dropped, to 10 percent in early 2020, and in November 2021, once the company had, in his words, “a cash balance that exceeds its debt due and is making money,” he set it to zero. The zero was not where he began. It was reached, once the company funded itself.

The companies that carried a number

The same input, set for companies at the stage SpaceX is at now, young and raising money in public markets, has never been zero.

In his pre-IPO valuations he assigned Uber 5 percent, Lyft 10, WeWork 20, Airbnb 10, and Zomato 10. The reasons he wrote beside them turn on the same thing: Lyft “dependent on external capital providers to stay a going concern”; Uber a pricing collapse that “leads to capital being cut off”; WeWork a debt load that leaves it exposed to “shocks in the real estate market.” Two of these sit closest to SpaceX. Airbnb was the soundest of the group: near profitability, little debt, a known brand. It still drew 10 percent, because it remained “a young, money losing company.” Zomato held a large cash balance, larger after its IPO, and still drew 10 percent; the cash pushed down the risk, he wrote, but the company “will need access to capital in future years to continue to survive.” A strong position did not earn a zero in his hands, and neither did cash on the balance sheet. Self-funding did.

Where the filing puts SpaceX

In the observed practice, the probability falls to zero once a company funds itself. The filing says SpaceX has not reached that point. It states that its access to the capital markets or other financing “may be adversely affected by factors beyond our control,” that it “may issue a significant amount of equity,” and that its liquidity is sufficient “for at least the next twelve months.” By the measures used in prior valuations — young, raising, dependent on outside capital — SpaceX sits with the companies that carried a number. This particular valuation carries zero.

Adding it up

The zero reaches most of the resulting valuation. More than four-fifths of the value lies beyond the ten-year forecast, in the terminal value, and the company has to survive the whole way to collect it. The cash requirement runs the other direction: near-term, inside the forecast, the commitments the earlier notes totaled against the raise. So the shape of it is a near-term reliance on raising capital, crossed in order to reach a value that is almost entirely back-loaded, with the one input that would price the survival risk left at zero. In the settings, where a failure costs half of fair value and is applied once, each percentage point of failure probability would subtract about $6.1 billion from the value of its operating assets. At zero it subtracts nothing, and the half recovery it would soften never comes into play.

Between April and June, Professor Damodaran sharpened and updated the valuation: he confirmed much of it and shifted more of its weight onto AI, doubling the AI revenue target. Through both passes, this one input stayed at zero. The model values the company at $1.25 to $1.35 trillion, with the offering at 138 percent of that. He asked his readers to check the numbers. We checked the one left at zero.

Whence the cash. Still open.

Part of a continuing read of the SpaceX offering documents, taken one at a time and in a single currency. The earlier notes totaled the disclosed cash requirement against the raise, read the two compute contracts against their terms, and tested the AI revenue assumption. This one checks the input the leading valuation leaves at zero.

On the comparison set: it is drawn from young, cash-using companies going public, the cohort that matches SpaceX. Professor Damodaran’s pre-IPO valuations since (Instacart, which had turned profitable on its advertising revenue, and Birkenstock, long profitable, both in 2023) were of cash-generating businesses, a different case.

The assumption

On June 4, Aswath Damodaran published his post-prospectus valuation, the work we would send a reader to first. On his numbers the equity is worth $1.25 to $1.35 trillion, and the offering price stands at 138% of it. In reaching it he made one change he flagged above the rest. He doubled his target for the AI business: “I will double my target revenues for AI from $80 billion to $160 billion.” He raised it on the company’s stated ambition, the Cursor acquisition (the enterprise AI coding company it holds an option to buy) and “the indications in the prospectus,” while noting he had “not much to go on” on enterprise revenue, and that the filing’s “focus ... is more on the increase in compute capacity ... than it is on revenues.”

He then named the risk he weighed most: “the one that I would be concerned about the most is that it will overreach in the AI business, beginning with an overestimate of the target market for AI products and services and the strength of its own competitive position in that market.” And he set the task for the reader: “listen but check the numbers for plausibility and make your own judgments.”

That is the assumption. The AI revenue line, doubled, named as the one to watch, handed to the reader to test.

The headline

The summary supplies what looks like the answer. Two compute agreements, both disclosed within the month. Anthropic, signed May 3, at $1.25 billion a month through May 2029, roughly $45 billion across the term. Google, disclosed June 5 in a free writing prospectus, at $920 million a month from October 2026 through June 2029, about $30.4 billion. Together near $75 billion of AI compute revenue, contracted and named. Read at the headline, that carries the doubled assumption. Close to half of the year-ten AI target, in signed contracts, inside a single month.

The release

The instruction was to read the rest, and the same documents price the commitment behind the headline.

Both agreements are terminable by either party on 90 days’ notice. Anthropic’s runs from signing. Google’s holds until after December 31, 2026, and its first dollar is contingent on SpaceX delivering 110,000 GPUs by September 30, with Google free before then to terminate or take fewer. Counting each notice period at the full rate, with delivery met, the most the disclosed terms can require is near $9 billion. The firm figure sits below that. The balance of the $75 billion is the part the customer may decline, and the part the filings price “at a reduced fee” with no amount stated.

The capacity reads in the same register. Colossus and Colossus II “collectively provide approximately 1.0 gigawatt,” and that cluster is at work: “Grok 5 ... is currently being trained at COLOSSUS II.” The leasing is described as selling “excess capacity ... to a limited number of third parties,” under a “dual monetization strategy,” with the company “permitting reallocation of the capacity for our own internal initiatives if needed in the future.” The next build carries the same claim: “the next phase of expansion at COLOSSUS II is designed to train our next-generation Grok 5.” The third-party revenue is the residual on a cluster whose first claim is the company’s own model.

There is a second reading the release supplies. Damodaran’s doubled figure is for “AI products and services from businesses,” priced in his model at a 25% operating margin. The two contracts are compute rental, the company as landlord, monetizing capacity it has by its own account in surplus. The headline that meets the products assumption comes from a different line than the one the assumption prices.

The counterparty

The release names one thing the summary does not. Google is among the “existing investors” party to the company’s Investors’ Rights Agreement, and Donald Harrison, President of Global Partnerships and Corporate Development at Google, “has served on our board since February 2015.” One of the two named contracts, the Google agreement, is with a shareholder represented on the board. The contract sits in the free writing prospectus; the ownership and the board seat sit in the related-person section and the director list of the May 20 prospectus. Each is disclosed, in a different place, in a different document.

Adding it up

That is what the release returns, read against the assumption the reader was asked to test. The AI line was doubled and named as the one to watch. The headline that appears to confirm it totals near $75 billion. The terms behind it commit a cancellable fraction, on capacity the filing calls excess and recallable, from a segment that is the company’s own, with one of the two held by an owner.

Every one of those facts is filed. The headline stands alone in the summary; the commitment, the capacity, and the relationship sit apart, in the risk factors, the subsequent-events note, the related-person section, and a separate free writing prospectus. The answer is in the documents and assembled in none of them. It reaches the reader who reads the entire prospectus and checks the numbers for plausibility, which is what the filing asked and what the analyst asked.

We did the reading and the arithmetic. Tested at the headline, the assumption holds. Tested at the terms, it waits on a delivery, a notice period, and capacity the company may take back.

Part of a continuing read of the SpaceX offering documents, taken one at a time and in a single currency. The earlier notes totaled the disclosed cash requirement against the raise, and read the two compute contracts against their terms. This one tests the AI revenue assumption the prospectus and its leading valuation both put forward.

The announcement

On June 5, 2026, SpaceX filed a free writing prospectus, a Rule 433 supplement to its IPO registration statement, disclosing a Cloud Service Agreement signed that day with Google LLC. In the filing’s words, Google “agreed to pay us $920 million per month from October 2026 through June 2029” for access to roughly 110,000 NVIDIA GPUs and supporting infrastructure, about $30.4 billion across the 33-month term. Capacity ramps through September “at a reduced fee.” If SpaceX fails to deliver the committed GPUs by September 30, 2026, then after a one-month grace period Google may terminate, or take fewer chips at a pro-rata-reduced fee. After December 31, 2026, either party may terminate on 90 days’ notice.

It is the second contract of this shape. On May 3, SpaceX signed cloud services agreements with Anthropic: $1.25 billion per month through May 2029, capacity ramping in May and June “at a reduced fee,” terminable by either party on 90 days’ notice, for roughly 325,000 GPUs across Colossus and Colossus II.

The actual commitment

Both contracts headline large. Each thins at every step when read through three figures: the full-term total, the part the customer cannot cancel, and the part the filing actually prices.

The cancellation terms set the ceiling, so take the most the disclosed terms can require, valued at the full rate. Google cannot terminate until after December 31, 2026, so the most it owes is October through the 90-day notice period that follows: about six months, roughly $5.5 billion, and even that is contingent on SpaceX standing up 110,000 GPUs by September 30. Before delivery, it is zero. Anthropic, terminable by either party on 90 days’ notice from signing, owes at most the notice period: about three months, or $3.8 billion if the reduced-fee ramp is counted at the full rate. These are generous ceilings. The real figures sit below them.

The term is the filing’s: terminable by either party on 90 days’ notice. The controlling shareholder has supplied the reason for it, in public. On May 28, Elon Musk wrote that “SpaceX has not committed to leasing Colossus for years,” that the short term “was our request, not Anthropic’s,” and that “if compute gets super tight,” SpaceX “might need it back.” The out reads less as a concession to the customer than as the company keeping capacity it may need for itself. The conditionality runs in both directions at once. The revenue depends on building the capacity, and the commitment depends on not needing it back.

And the firm windows are the ramp months, priced in both contracts at “a reduced fee” with no amount and no floor stated. The full monthly rates, $920 million and $1.25 billion, describe the periods after the ramp, which the customer may never reach. The committed period is the period the filings decline to price.

So of roughly $75 billion in combined headline, the most the disclosed terms can require is about $9 billion: both notice periods, valued at the full rate, with SpaceX delivering. That is the ceiling, and the real figures sit below it. A fraction of the headline, either way.

The cash call

What the contracts do commit SpaceX to is the build. And the build comes first.

The filing states one hard capacity figure: Colossus and Colossus II “collectively provide approximately 1.0 gigawatt of compute power,” reached in the first quarter of 2026. The two contracts call for roughly 435,000 GPUs across those facilities, and xAI trains its own models on the same hardware. If the existing gigawatt is largely committed, much of it to Anthropic and the rest to xAI’s own training, then Google’s 110,000 GPUs are new capacity. If some of it is free, the new build is smaller. Either way, the capacity has to exist, and be running, by September 30.

The cost of it is in the company’s own numbers. The AI segment spent $12.7 billion of capital expenditure in 2025 and $7.7 billion in the first quarter of 2026, roughly $20 billion to stand up about a gigawatt. Give the company every benefit on the economics: call the per-gigawatt cost lower than that, assume it falls with scale. None of that is the argument here. The argument is only how much has to be spent, and when.

Even at a discount, Google’s 110,000 GPUs — about a fifth of a gigawatt — are several billion dollars of capital, and the deadline is September 30. It is June. The spend goes out this quarter and next, on top of the $7.7 billion the AI segment already laid out in the first quarter, to bring a new cluster of power and cooling online inside four months. The company says, in the prospectus, that it brought a Colossus II cluster online in 91 days. From June, 91 days reaches September, at the edge of the date.

That is the sequence the contract sets. The capital goes out now, this quarter, to build capacity that earns a monthly fee beginning in October, at a reduced rate through the ramp, under an agreement the customer can leave after December. The cash leads the revenue by the full cost of the build, and the revenue at the end of it is cancellable and, where firm, unpriced.

What it adds to

Yesterday’s piece found a cash requirement the offering does not meet, disclosed but not sized, and expected the largest of it to land in the near term, ahead of revenue. The contracts are consistent with exactly that. Dated to a September delivery and an October start, they put a name and a date on the near-term spend the cash analysis anticipated.

They do not, on their own, enlarge the gap, and we do not claim they do. They sharpen its timing. The build precedes the revenue, the revenue is mostly optional, and where the cash comes from is the question the last piece left open. Nothing filed this week has closed it.

A note on value

These contracts have also been read as a verdict on the company’s worth: proof, in some accounts, that the cash question is now answered. Settling the valuation is not this piece’s purpose, and it does not try. But the question earns a note, because the leading public valuation of SpaceX rests, at one point, on the financing question these contracts make concrete.

Aswath Damodaran’s post-prospectus valuation is a well-considered and excellent model, the most transparent public assessment of the company, every input visible, the spreadsheet posted for anyone to download and change. It is the work we would send a reader to first. On his own numbers he arrives at $1.25 to $1.35 trillion of equity value, and he notes that the offering price stands at 138% of it.

He asked his readers to do one thing, and we did it. “Listen,” he wrote of the sales pitches to come, “but check the numbers for plausibility and make your own judgments.” So we checked one. His model carries an explicit input: a probability of failure of zero. For a company that burns cash and needs capital well beyond the offering, a zero is a strong number, the kind worth confirming before it is relied on.

The documents say the same thing, in their own register. The prospectus describes this company’s funding in the conditional throughout: liquidity “sufficient for at least the next twelve months,” an intention that it “may issue a significant amount of equity,” funding that “depends on continued access to the capital markets,” and the compute contracts themselves, terminable on notice and contingent on a delivery the company “may fail” to make. That is the language of a risk to be confirmed before it is assumed away.

We note only what was asked, and what we found: an excellent model rests, at this one input, on a zero, and the filing that supplies its other inputs describes the same risk in “may.”

The contract filed this afternoon

On June 5, 2026, the afternoon this was published, SpaceX filed a free writing prospectus under Rule 433. It confirmed the June 3 pricing amendment and disclosed something new: a Cloud Service Agreement signed that day with Google LLC. Google agreed to pay $920 million a month, from October 2026 through June 2029, for access to roughly 110,000 NVIDIA GPUs and supporting infrastructure — about $30.4 billion over the term. Capacity ramps through September at a reduced fee. If SpaceX fails to deliver the committed GPUs by September 30, 2026, then after a one-month grace period Google may terminate, or take fewer chips at a lower fee. After December 31, 2026, either party may leave on ninety days’ notice. Google keeps its own models and data.

To earn the first dollar in October, the capacity has to exist by September 30 — 110,000 GPUs, standing and running, four months from now. That is capital deployed now, against revenue collected monthly, later, under a contract the customer can cancel. The contract sets out the revenue. The capital to build the capacity behind it appears nowhere. One filing, the whole question.

A narrower question

Set aside the familiar objections — that SpaceX loses money, burns cash, trades at a high multiple of revenue. They are weak against a young company, and they have been answered. The question here is narrower, and it turns on amounts and dates. Allow the company its growth and the valuation, any of the published ones, at their most generous: the cash the disclosed plan requires, on the schedule the disclosures set, still runs past what the offering provides. What would close the difference is disclosed as a possibility, in a single sentence.

We estimated, then; we now estimate

On May 10 we built a consolidated cash model for 2025 through 2030. Its headline was a cumulative shortfall near $165 billion, against an IPO then expected to net the company roughly $50 billion. Most of that model was estimate. It is now, in large part, record.

The 2025 column has been replaced with the company’s own numbers. Revenue came in at $18.7 billion, not the $15 billion estimated. Operating cash flow was $6.8 billion, which the filing attributes largely to Starlink subscriptions billed in advance. Capital expenditure was $20.7 billion, against a $15 billion estimate. Free cash flow, operating cash less capex, was negative $13.9 billion — within a tenth of a billion of the 2025 free cash flow Goldman Sachs is reported to carry. A subtraction anyone can run off the cash-flow statement and the lead underwriter’s own model reach the same figure by separate routes. The proxy and the bank agree.

The IPO line firmed, and in the company’s favor. Priced at $135, with the over-allotment, the offering grosses about $86 billion. The underwriters — Goldman, Morgan Stanley, and three others — agreed to a fee under 0.75%, a few hundred million on the largest IPO ever attempted. After the fee and the contractually required repayment of the $20 billion bridge loan, about $65 billion reaches operations. Add the cash already on the balance sheet and the company has roughly $89 billion to work with.

The first quarter of 2026 is in the filing as well. Consolidated capital expenditure was $10.1 billion in three months, $7.7 billion of it in the AI segment alone. At that quarterly pace the model’s later years look conservative, though a single quarter is thin evidence for a trend.

Update only what is now known — the actual 2025, the firmed offering — and the model’s cumulative gap is about $159 billion. It came down from the May estimate, because the raise firmed larger than expected. And it is still $159 billion: roughly twice the largest initial public offering ever priced. After that offering.

These are different measures. The $235 billion of the earlier piece was the sum of the commitments themselves — what must be spent. The $159 billion here is the cumulative cash position the model projects after the offering and after operations — what is left unfunded.

The valuations, low to high

Since May the major valuations have published, and they span more than a trillion dollars. Morningstar, the most bearish, puts fair value at $780 billion — under half the offering price — and rests the case on the AI business and the governance, not on the cash. Aswath Damodaran, the most-cited independent valuation, arrives at $1.25 to $1.35 trillion — below the price, which his own spreadsheet marks at 138% of his value. The underwriters reach higher. Morgan Stanley, a co-lead, reportedly models $330 billion of revenue by 2030 and $3.4 trillion by 2040. Goldman, the lead, projects $474 billion of revenue by 2030, with negative free cash flow every year through 2030 and the first positive year in 2031.

The numbers disagree by more than a trillion dollars. On one point they are identical. Each values the business; none sizes the cash required to build it. Goldman’s coverage goes furthest: it carries negative free cash flow every year through 2030 and names “future equity raises,” without quantifying them. The valuation is the aspiration, and the aspiration lives in the addressable market: $28 trillion in the prospectus, $26 trillion of it AI, a figure Damodaran himself says “borders on fantasy.” The larger the market assumed, the more capacity the company must build to reach it, and the more cash that build requires. The valuation and the cash requirement are the same quantity, seen from two ends.

Fund the gap, or lower the aspiration

Here is the trade, on the most favorable assumptions.

To hold the aspiration — the trillion-and-a-quarter-to-1.8-trillion business — the company has to build what the revenue assumes: the Colossus II capacity behind the Google and Anthropic contracts, the Starlink and Starship deployment, and, if it is built, Terafab. The chip plant alone is $55 to $119 billion. The S-1 disavows any commitment to it — “neither Tesla nor Intel are obligated to remain a part of the project,” an obligation that runs both ways — even as, on June 3, the day the offering firmed, Grimes County approved a 100% tax abatement for the plant, against SpaceX’s pledge of a $5 billion minimum investment and 1,800 jobs.

On the cheapest version, by our model, the disclosed plan still ends 2030 about $135 billion short of the offering plus the cash on hand; build the plant and it runs toward a quarter-trillion. The difference comes from new equity, which dilutes — and it would dilute from the offering price, a level the leading valuations already sit below, raising capital above the value their own work assigns before a single new share changes hands.

Decline to build it — take “not committed” at its word — and the trade reverses. No new capital is required. But the capacity is not built in time: no 110,000 GPUs for Google by September, no wafers for 2030. The contracted revenue doesn’t arrive, and the business that remains is smaller than the published valuations assume.

There is no third cell. Nothing pairs a high valuation with a low cash requirement, because the valuation is a claim about a business large enough to need the cash. Fund the gap, or lower the aspiration. The published valuations assume the first and price neither.

Where the cash comes from

It is a fair question to ask of any cash-burning company, and the honest answer is the one everyone has already given: more will be needed, and more will be available. The market has funded SpaceX privately for twenty years. In 2025 the company raised $26.35 billion through financing activities — roughly $16 billion of AI-segment borrowing, the balance from stock sales — to cover a year in which it spent $20.7 billion against $6.8 billion of operating cash. That figure comes from the cash-flow statement itself. The access is real, and the assumption that it continues is reasonable on its face.

It is the scale that changes the arithmetic. Damodaran’s model implies an outside-capital need near $13 billion — the cumulative trough his cash flows fall into before they turn, which the IPO covers several times over — with the probability of failure set to zero. His text, separately, names overreach in AI and “tens of billions more in capital expenditures” as his chief concern. At $13 billion the cash assumption is invisible. At $135 billion and up it becomes the question, because a raise that size can be reflexive: it would come before the revenue, and large pre-revenue raises can pull the price down as they go.

The customers did not finance it. A customer that wanted the capacity could have prepaid, or taken equity to fund the build — a common structure. Google did neither; it pays monthly, in arrears, at a reduced rate during the ramp, and it can leave after December. Anthropic, the same. The two anchor customers, together about $26 billion a year, financed none of the construction. The build is SpaceX’s to carry.

The prospectus addresses its own liquidity in the conventional way. It vouches for “at least the next twelve months,” and it names the two doors without sizing either: it may “reduce future capital expenditures in this segment and reallocate,” or it may “raise additional capital or seek alternative financing sources.” Twelve months, and a choice. The horizon every valuation runs on is five years and more.

Adding it up

That is the whole of it. Choose your valuation, from Morningstar’s $780 billion to the underwriters’ multiples of it. Choose your assumptions about efficiency, about access to the capital markets, about the size of the market. Even at the company’s own stated positions, a near-term cash requirement remains that the offering does not meet. The model does only the arithmetic the filings invite, and stops there.

The requirement is disclosed, but not quantified. What that leaves is the follow-on question — from where — answered only by the language added in the amendment, as the closing line of the risk factor on integrating acquisitions, the paragraph that names xAI, the spectrum, Terafab, and Cursor:

“We may issue a significant amount of equity in connection with future transactions.”

It is the entire disclosure of the largest unfunded requirement in the document. The transactions it would fund are, by the same filing, already signed. Everyone added it up — the company, its bankers, its analysts — and the disclosure that resulted is that one sentence, which in this disclosure environment may be all that is required.

Whence the cash. Still open.

The New York Times piece on Tuesday, May 26, along with other journalistic coverage of the S-1’s compensation disclosures, read the 2026 SpaceX grants as a departure from corporate governance conventions, with the 2025 Tesla CEO Performance Award as the reference baseline. That reading is insightful on its own, and the analysis bears further fruit when compared across time frames and entities.

Tesla’s 2018 CEO Performance Award, approved by shareholders in March 2018. Tesla’s 2025 CEO Performance Award, approved by shareholders in November 2025. SpaceX’s 2026 grants to Mr. Musk, awarded in January 2026 and March 2026, disclosed in the S-1 registration statement filed May 20, 2026.

Each iteration carries features the others do not. The sequential reading across three packages, three legal frameworks, and three windows of corporate scale shows motion across the architecture’s compensation dimension.

Historical Precedent: The 2018 Tesla CEO Performance Award

The Tesla board approved the 2018 CEO Performance Award on January 21, 2018. Tesla shareholders approved the award in March 2018 with 73% support from disinterested shareholders.

The award consisted of 12 tranches of performance-based stock options with an exercise price of $350.02. Each tranche required achievement of one market capitalization milestone paired with one operational milestone. Market capitalization milestones began at $100 billion and increased in $50 billion increments to $650 billion. Operational milestones consisted of 16 thresholds — 8 revenue targets and 8 Adjusted EBITDA targets. Any operational milestone could be paired with any market capitalization milestone; each milestone could be used only once. The grant date fair value was estimated at approximately $2.3 billion; the maximum potential value at full vesting was approximately $55.8 billion at grant.

Tesla was incorporated in Delaware at the time of grant. The award was subject to Delaware corporate law.

Mr. Musk received no salary, cash bonus, or other equity compensation during the award’s term. The package was designed as 100% at-risk performance compensation.

A shareholder derivative action filed in 2018, Tornetta v. Musk, challenged the award. The Delaware Court of Chancery ordered rescission of the award in January 2024 and rejected a subsequent ratification attempt in December 2024. The Delaware Supreme Court reversed the rescission in December 2025, and the 2018 award was reinstated.

Current Framework: The 2025 Tesla CEO Performance Award

The Tesla board approved the 2025 CEO Performance Award on September 3, 2025, following review by a Special Committee of disinterested directors formed on January 8, 2025. Tesla shareholders approved the award at the annual meeting in Austin, Texas on November 6, 2025 with 75% of voting shares in favor.

The award consists of 12 tranches of performance-based restricted stock covering 423,743,904 shares, representing approximately 12% of Tesla’s outstanding stock. Each tranche requires achievement of one market capitalization milestone paired with the cumulative achievement of a required number of operational milestones. Market capitalization milestones range from $2 trillion to $8.5 trillion across the 12 tranches, with $500 billion increments through tranche 10 (reaching $6.5 trillion) and $1 trillion increments for the final two tranches ($7.5 trillion at tranche 11 and $8.5 trillion at tranche 12). Each market capitalization milestone is determined using both a 30-day and a 6-month average, requiring the level to be sustained rather than transiently touched. Operational milestones consist of 12 thresholds: 1 million Robotaxis in commercial operation, 1 million AI Bots delivered, 10 million active FSD subscriptions, 20 million Tesla Vehicles delivered, and eight Adjusted EBITDA thresholds from $50 billion to $400 billion (with the $400 billion threshold available three times across non-overlapping four-quarter periods). Each tranche N requires achievement of any N of the 12 operational milestones. Achievement of the eleventh and twelfth tranches additionally requires development of a CEO succession plan approved by the Administrator. The maximum potential value at full vesting is approximately $1 trillion; the preliminary grant date fair value was estimated at approximately $87.75 billion.

Tesla reincorporated from Delaware to Texas on June 13, 2024. The 2025 award is subject to Texas corporate law and any derivative litigation would be decided under Texas standards.

The 2025 award is structured as restricted stock rather than stock options. Mr. Musk’s exercise of voting rights on the restricted shares is governed by a separate Voting Agreement entered into between Mr. Musk and Tesla, attached to the award agreement as Annex I, until shares become Earned Shares.

Section IV of the award agreement, titled “Determination of Earned Shares upon Change in Control,” provides that “Notwithstanding Sections I, II and III above, in the event of a Change in Control, the Operational Milestones shall be disregarded and the Market Capitalization shall equal the product of (a) the total number of outstanding Shares immediately prior to the effective time of such Change in Control... and (b) the greater of (i) the most recent closing price per Share immediately prior to the effective time of such Change in Control... and (ii) the per Share price (plus the per Share value of any other consideration) received by the Company’s shareholders in the Change in Control.” The “Notwithstanding” lead establishes that Section IV overrides the operational and market capitalization milestone provisions in Sections I, II, and III. In a Change in Control, the operational milestones are disregarded and market capitalization is calculated using the transaction price. The dual 30-day and 6-month averaging requirements that otherwise apply to milestone achievement are also displaced; the transaction price itself satisfies the market capitalization calculation at the moment of the Change in Control, without any sustaining requirement. If a Change in Control occurs at a transaction price implying $8.5 trillion in total Tesla market capitalization, the operational milestones become irrelevant, the sustaining requirement does not apply, and all 12 tranches vest at that implied valuation.

Mr. Musk continues to receive no salary, cash bonus, or other equity compensation during the award’s term.

Further Framework: The 2026 SpaceX Grants

The SpaceX board approved a performance-based restricted stock grant to Mr. Musk on January 13, 2026. Following the company’s 5-for-1 forward stock split on May 4, 2026, the grant covers 1 billion shares of Class B common stock. The grant vests upon achievement of specified market capitalization milestones across 15 tranches and the Company’s establishment of a permanent human colony on Mars with at least one million inhabitants. Both the applicable market capitalization milestone and the human colony milestone must be met for a tranche to vest. As originally granted, market capitalization milestones began at $400 billion and increased in $400 billion increments to $6 trillion. Following the xAI Merger that closed on February 2, 2026, the milestones were equitably adjusted under Section 3 of the award agreement to a range of $500 billion to $7.5 trillion in $500 billion increments.

On March 23, 2026, the SpaceX board approved a second grant of 302 million Class B restricted shares (post-stock-split), replacing a performance award originally granted by xAI to Mr. Musk on November 26, 2025. This award vests upon achievement of specified market capitalization milestones across 12 tranches and the Company’s completion of non-Earth-based data centers capable of delivering 100 terawatts of compute per year. Market capitalization milestones range from $1.065 trillion to $6.565 trillion, with each milestone reflecting $500 billion in additional valuation. Both the applicable market capitalization milestone and the orbital data center milestone must be met for a tranche to vest.

The combined grants total approximately 1.3 billion Class B restricted shares.

Class B common stock carries 10 votes per share, compared to one vote per share for Class A common stock. Section 6(a) of both award agreements provides that “Participant shall have all the rights and privileges of a holder of Class B Common Stock in respect of the Restricted Shares that have not been forfeited, including the right to vote the Restricted Shares from the Date of Grant.” Footnote (1) of the principal shareholders table in the S-1 confirms that the 1,302,072,285 restricted Class B shares “may be voted by Mr. Musk.” The voting power conferred by these shares is included in Mr. Musk’s 85.1% combined voting power disclosed in the S-1.

SpaceX was incorporated in Delaware on March 14, 2002 and reincorporated as a Texas corporation on February 14, 2024. The 2026 grants are subject to Texas corporate law. Under the controlled-company exemption SpaceX has elected, the compensation and nominating committee is not required to consist of independent directors.

The S-1 states that no executive officer is party to an employment agreement or severance arrangement providing for payments or benefits upon termination of employment or a change in control of the Company. No equity award agreements provide for benefits upon termination of employment or change in control. The 2026 award agreements do, however, provide expressly for modification authority. Section 3 of each award agreement authorizes the Board (or a designated committee thereof) to equitably adjust the Market Capitalization Milestone in the event of material corporate actions, acquisitions, or dispositions. Section 14 authorizes the Company to amend the agreement unilaterally for amendments that do not adversely affect Mr. Musk’s rights, and to amend with Mr. Musk’s written consent for amendments that do. Section 16 grants the Board final authority to interpret the agreement.

The grant date fair value of the SpaceX CEO Award was determined to be $90.40 to $95.92 per share for each tranche; the grant date fair value of the AI CEO Award was determined to be $91.47 to $95.92 per share for each tranche, calculated using a Monte Carlo simulation with an expected term of 30 years (the midpoint of a 10-to-50-year assumed range), expected volatility of 45%, and a risk-free interest rate of 4.91%. As of March 31, 2026, the company has classified both performance milestones as improbable for accounting purposes, and no share-based compensation expense has been recognized.

Mr. Musk receives a base salary of $54,080 and no cash bonuses.

The Pattern

The same procedural mechanisms — committee approvals, shareholder votes, governing-law selections, voting agreements — appear across all three iterations, but the environments in which they operate have changed materially. Their constraint value has diminished as the iterations have progressed. Independent committee approval, disinterested shareholder votes, voting rights deferred until vesting, change-of-control vesting acceleration — these features are responses to external accountability that constrain compensation arrangements at companies without controlled-company structures. The architecture has reduced the operative external accountability such that these features are no longer functionally required. The packages reflect this.

The voting rights on unvested shares illustrate the pattern directly. Restricted shares are typically issued without unrestricted voting rights pre-vesting because the standard rationale for restricted stock is to require performance before conferring voting power. The 2026 SpaceX grants confer voting rights from the date of grant without restriction. Mr. Musk votes the 1.3 billion restricted Class B shares as part of his 85.1% combined voting power. The deferral that the standard rationale prescribes operates differently in an architecture where the controller already commands a majority of voting power; the deferral’s structural purpose is functionally absent.

One feature visible in the table merits attention. The 2025 Tesla award agreement, at Section IV, includes an explicit change-of-control bypass: in a Change in Control, the operational milestones are disregarded and market capitalization is determined using the transaction price. The 2018 Tesla award contained no equivalent provision and structurally could not have done so; Tesla had no controlled affiliate of sufficient scale to acquire it at the upper milestone valuations. The 2026 SpaceX grants contain no automatic change-of-control acceleration; the S-1 confirms that no equity award agreement provides for benefits upon change of control. The 2026 SpaceX grants do contain, however, expressly provided modification authority through Sections 3, 14, and 16 of the award agreements: equitable adjustment of the Market Capitalization Milestone, unilateral amendment authority, and final Board interpretive authority. The architecture that could effect outcomes similar to Section IV’s automatic mechanism is written into the 2026 SpaceX award agreements as discretionary modification rather than automatic vesting. The pattern across the three iterations is one of motion: a provision that could not have existed in 2018 Tesla, exists explicitly in 2025 Tesla, and can be effected through modification authority in 2026 SpaceX.

The Monte Carlo simulation disclosed for the 2026 SpaceX grants — expected term of 30 years (a midpoint of an assumed 10-to-50 year range), expected volatility of 45%, risk-free interest rate of 4.91% — applies standard option-pricing techniques to estimate the grants’ value at the time of grant. The simulation captures the grants as financial instruments with specified terms operating over a defined horizon. The simulation is an admirable attempt to quantify the unquantifiable, but it is not load-bearing. The packages’ actual operation depends on the architecture’s permissions, which include the Board’s discretionary equitable adjustment authority under Section 3 and the Company’s unilateral amendment authority under Section 14 of the award agreements. The simulation captures the form of the disclosure with precision; the underlying arrangement carries flexibility the simulation does not measure.

The current packages reflect the current state of the architecture’s compensation dimension. The structural features that distinguish the 2026 SpaceX grants from prior conventions also make subsequent modifications available within the architecture’s permissions.

What the Iteration Produces

The three iterations describe a pattern that has continued for eight years. Each iteration’s features became visible at the time of grant. Each iteration’s structural relationship to the prior iteration is visible in retrospect.

The 2026 grants may not be the final iteration. History suggests they will not be, and the agreements themselves contain the authority to modify them.

The comparison across iterations presents contrasts.

Documents and citations

This piece draws on the following publicly filed documents.

SpaceX — Form S-1 Registration Statement filed May 20, 2026, available at SEC EDGAR. Principal source for the 2026 grants, the corporate structure, the controlled-company exemption, and the voting power disclosures. Specific citations: Compensation Discussion and Analysis section (2026 Compensation Developments); footnote (1) of the Principal Shareholders table; Note 15 (Share-Based Compensation) of the financial statements.

SpaceX — Exhibit 10.6, Class B Restricted Stock Award Agreement, dated as of January 13, 2026, filed with the S-1.

SpaceX — Exhibit 10.7, Class B Restricted Stock Award Agreement, dated as of March 23, 2026, filed with the S-1.

Tesla — Form 8-K filed November 7, 2025, Exhibit 10.2, 2025 CEO Performance Award Agreement, available at SEC EDGAR. Principal source for the 2025 Tesla package’s structural terms, including Section IV (Determination of Earned Shares upon Change in Control) and Section V (Milestone Adjustments).

Tesla — Form DEF 14A filed January 2018, available at SEC EDGAR. Principal source for the 2018 Tesla CEO Performance Award’s structural terms and the March 2018 shareholder approval mechanism.

Tornetta v. Musk, 310 A.3d 430 (Del. Ch. 2024) (Tornetta I); Tornetta v. Musk, C.A. No. 2018-0408-KSJM (Del. Ch. Dec. 2, 2024) (Tornetta II); In re Tesla, Inc. Derivative Litigation, No. 534, 2024 (Del. Dec. 19, 2025) (Delaware Supreme Court reversal).

The New York Times — “SpaceX’s Unconventional Corporate Arrangements Favor Elon Musk,” published May 26, 2026.

This is the fourth piece in a sequence examining the SpaceX S-1 disclosure architecture. Prior pieces: Adding It Up — The $235 Billion Cash Gap; SpaceX’s $28,500,000,000,000 TAM; The Three Layer Cake: SpaceX’s Governance Structure.

Summary

SpaceX’s S-1 estimates the company’s quantifiable total addressable market at $28.5 trillion. The aspiration — described in the prospectus as “the largest actionable TAM in human history” — gets translated into a specific methodology that comprises seven components across three operating segments: $370 billion in Space, $1.6 trillion in Connectivity ($870 billion Broadband and $740 billion Mobile, plus enterprise and government), and $26.5 trillion in AI ($2.4 trillion infrastructure, $760 billion consumer subscriptions, $600 billion digital advertising, and $22.7 trillion enterprise applications). The aspiration and the disclosure are not the same thing.

The disclosure raises cash implications and addresses them qualitatively. The S-1’s Risk Factors acknowledge that TAM estimates “may prove to be inaccurate” and that the growth strategy “will require significant capital expenditures.” The disclosure architecture permits the issue to be named in risk-factor prose without being quantified. The framework’s analytical work supplies the quantification: at industry-standard capital intensity applied to operationally defensible share targets for the six methodology-bearing components, the implied capital base through 2030 is approximately $850–900 billion. Against disclosed funding sources of approximately $200–275 billion, the funding gap is approximately $575–700 billion.

The aspiration requires capital at scale. The disclosed funding architecture, dominated by Class A share issuance for major strategic commitments plus the IPO cash event, is structured for acquisitions and partnerships rather than for the physical infrastructure that the TAM construction implies. Stock-denominated consideration funds acquisitions where the seller accepts stock and partnerships where the counterparty accepts stock-denominated obligations. The physical infrastructure that the TAM construction implies is paid for in cash, ultimately sourced from operating cash flow, debt issuance, or equity issuance converted to cash in the capital markets. At the implied scale, the equity-issuance-to-cash channel would require capital markets to absorb dilution at a magnitude that itself becomes a constraint.

The disclosed governance and control structure provides controller discretion on the timing and cost of investments, separate from their executional risk. Mr. Musk holds approximately 85% of voting power through the dual-class share structure. The company operates under the controlled-company exemption. The controller decides when and on what terms capital gets deployed against the implied requirements. The executional risk — whether Starship works, whether orbital AI compute is feasible, whether Terafab succeeds, whether Starlink Mobile reaches the addressable device base — is separate from the discretionary control over the deployment.

Methodology and Anchoring

The headline TAM disclosure appears in the Business section under “Our Market Opportunity,” and reads in full:

“We believe we have identified the largest actionable total addressable market (’TAM’) in human history. We estimate that our quantifiable TAM is $28.5 trillion, consisting of $370 billion in Space from space-enabled solutions; $1.6 trillion in Connectivity across $870 billion in Starlink Broadband and $740 billion in Starlink Mobile as well as additional opportunities in enterprise and government; $26.5 trillion in AI across $2.4 trillion in AI infrastructure, $760 billion in consumer subscriptions, $600 billion in digital advertising, and $22.7 trillion in enterprise applications. For illustrative purposes of sizing our addressable market opportunity, we exclude China and Russia from our global estimates.” [S-1, Prospectus Summary, p. 8]

The Risk Factors section addresses the methodology in compact form:

“With respect to our Space segment, these estimates rely in part on estimates published by Novaspace regarding the size of the global market for space-enabled solutions, including spacecraft manufacturing, launch services and related activities. Our connectivity market estimates are based in part on estimates of the number of households, businesses, aircraft and maritime vessels globally derived from third-party sources, together with assumptions regarding ARPU and monthly service revenue derived from third-party industry data and our internal expectations regarding pricing, adoption rates and service penetration across different geographic regions and economic environments. Our AI market estimates are based in part on projections of global data center compute demand from third-party sources, including estimates published by RAND Corporation, together with internal assumptions regarding the portion of global compute capacity that may be utilized for AI workloads and other operational assumptions such as power usage, utilization rates and pricing.” [S-1, Risk Factors, p. 50]

The Industry and Market Data section lists 41 numbered citations from sources including Novaspace, RAND Corporation, Silicon Data, Omdia, Ericsson, Euromonitor International, the Digital Cooperation Organization, McKinsey & Company, Boston Consulting Group, JLL, the International Energy Agency, the World Bank, and others [S-1, pp. iii–iv]. The S-1 notes:

“Some market data and statistical information contained in this prospectus are also based on management’s estimates and calculations, which are derived from our review and interpretation of publicly available industry publications, our internal research and our knowledge of the markets in which we currently, and will in the future, operate, as well as the sources referred to above. This information involves a number of assumptions and limitations, and you are cautioned not to give undue weight to such information.” [S-1, General Information, p. iv]

The Risk Factors include the company’s own acknowledgment that the TAM estimates may be inaccurate: “The estimates of future market opportunity and forecasts of market growth, and our ability to capture such markets, included in this prospectus may prove to be inaccurate.” [S-1, Risk Factors summary, p. 21]

On the anchor year. The TAM disclosure does not specify a year. Industry analytical convention requires market-size forecasts to anchor to a specified horizon — McKinsey publishes “AI Market 2030”; Gartner publishes annual forecasts with explicit projection years; Bain anchors strategic studies to specified horizons. The S-1’s TAM is presented in present tense, “our quantifiable TAM is $28.5 trillion,” without a year. This is a methodological gap. The construction itself synthesizes inputs of different vintages — present-year measurements (the DCO digital economy 2026; the World Bank GDP 2026), forward-projected demand (RAND power demand to 2030), forward-projected pricing held from 2025 (the Silicon Data H100 rental rate), and population counts that change slowly. The aggregation does not correspond to any specific year.

For analytical purposes this analysis anchors to 2030. This is the conventional analytical horizon, the period to which most cited sources project, and a conservative choice — 2030 is approximately the nearest defensible horizon for a forward-looking TAM, and most projection sources extend through 2030 with declining confidence past that point. The S-1 does not anchor; the framework imposes the anchor for analytical purposes and notes the gap.

Caption: The disclosed TAM construction is dominated by a single line — Enterprise Applications at $22.7 trillion — that derives from a measurement of the global digital economy rather than the conventional enterprise applications market. The remaining six components, taken together, account for $5.7 trillion (20% of the total) and bear methodologies that triangulate against named industry sources.

Space — $370 Billion

The Space TAM is constructed from a single source: Novaspace’s 12th Edition Space Economy Report, January 29, 2026 [S-1 citation xxi]. The Risk Factors paragraph names the construction directly: estimates from Novaspace covering “spacecraft manufacturing, launch services and related activities.”

Novaspace, formed in 2024 from the merger of Euroconsult and SpaceTec Partners, is among the most cited specialist research firms for the space industry. Their published 2024 global space economy estimate is approximately $596 billion. Their medium scenario projects the commercial space economy reaching approximately $700–800 billion by 2030 and $944 billion by 2033, with a long-range scenario extending to $1.8 trillion by 2035.

The Space Foundation’s “The Space Report 2025 Q2” [S-1 citation xxxii] places the 2024 global space economy at $613 billion, split approximately $293 billion commercial and $320 billion government. The $370 billion SpaceX-addressable figure sits between Novaspace’s current commercial figure and the Space Foundation’s combined total — consistent with a 2030 horizon for the commercial space economy at approximately 45–50% addressable share for SpaceX.

The operational reality: SpaceX’s Space segment generated $4,086 million in 2025 revenue and $619 million in Q1 2026 (approximately $2.5 billion annualized), with a Q1 2026 operating loss of $662 million [S-1, Segment Operating and Financial Data, p. 23]. Mass to orbit reached 7,400 metric tons cumulatively, with 99% mission success across the Falcon family [S-1, Business — Our Unparalleled Launch Capabilities, p. 14]. Federal customers represent the majority of Space segment revenue: “In 2025, we launched 11 of 12 National Security Space Launch (NSSL) medium and heavy lift missions and all five U.S. crew and cargo missions to the International Space Station for NASA” [S-1, Business — Our Leading Capabilities Across Space, Connectivity, and AI, p. 16].

The Space line is the most defensible component of the TAM disclosure. The methodology cites a credible specialist source; the source’s own projections support a figure in this range for 2030 commercial space economy; SpaceX is operationally positioned through launch dominance and vertical integration to address a meaningful share. The Space line is also the regulatory-compliant articulation of the company’s broader positioning across multiple disclosure registers — what the company has expressed less formally elsewhere as the intent to define the commercial space economy.

The capital implication. Space is the most capital-intensive segment per the S-1’s own disclosure. 2025 Space capex was $3,832 million against $4,086 million of segment revenue — 94% capex-to-revenue, with Q1 2026 at 170% [S-1, Capital Expenditures, p. 23]. Of 2025 capex, $3,004 million funded Starship development alone, approximately 75% of total Space capex [S-1, Prospectus Summary, p. 5]. The Space segment is fundamentally a development-stage business; the existing capital deployment is the down payment on Starship reaching commercial viability at scale.

Industry-standard capital intensity for vertically integrated space companies addressing multiple market segments is approximately 3.0–4.0x PP&E-to-revenue at maturity, based on benchmarks from aerospace prime contractors and satellite operators. Applied to $370 billion in addressable Space revenue, the implied PP&E base at full addressable scale is approximately $1.1–1.5 trillion. The operationally relevant question is what capital is required through 2030 to position the company for that scale: cumulative Starship development through commercial readiness (estimated $50–75 billion by industry analysts), satellite manufacturing infrastructure expansion ($20–30 billion), launch infrastructure across multiple sites ($15–25 billion), and supporting infrastructure including propellant production and ground systems ($10–20 billion). Cumulative incremental Space capital required through 2030 to operate at the scale implied by the TAM positioning: approximately $110–165 billion beyond current PP&E.

The Space segment is loss-making at current scale. Operating cash flow from Space is negative; capital must come from sources outside the segment. The S-1 confirms approximately one-fifth of total company revenue is federal [S-1, Risk Factors, p. 57]; the Space segment skew is much higher, with federal customers representing the majority of segment revenue. The capital required to position SpaceX to address the $370 billion claimed Space market through 2030 is financeable only with federal procurement that expands at scale, with continued cross-segment subsidization from the Connectivity segment’s profitability, with external financing — or with all three.

Connectivity Broadband — $870 Billion

The Connectivity Broadband construction follows the methodology described in the TAM Risk Factor paragraph: estimates of household counts from third-party sources, multiplied by ARPU and monthly service revenue assumptions. The named third-party sources include Omdia’s “Broadband Op Subs by Technology” [S-1 citation xxiii], Ericsson’s “Global Fixed Broadband Market Outlook” [S-1 citation v], Grand View Research’s broadband services market analysis [S-1 citation viii], Euromonitor International’s households-by-country data [S-1 citation vi], and IDC’s “Consumer Market Model H2 2025” [S-1 citation ix].

The reproducible math: approximately 2.3 billion households globally per Euromonitor’s 2026 edition; excluding China (approximately 480 million households) and Russia (approximately 55 million households) leaves approximately 1.75 billion addressable households. At a weighted-average ARPU of approximately $42 per month, 1.75 billion households × $42 × 12 months produces $882 billion, within rounding of the $870 billion disclosed. The construction is reproducible from the cited inputs.

The triangulation point: Omdia’s 2030 global fixed broadband revenue projection is approximately $440–470 billion annually for the entire global fixed broadband market — terrestrial and satellite combined. Ericsson’s projection is similar at $400–450 billion. Grand View Research extends to $650–700 billion. The SpaceX-claimed $870 billion addressable broadband market is approximately 2x the named sources’ projections for the entire global fixed broadband market in 2030, including all terrestrial operators.

The risk factor language is precise. Counts come from the third-party sources. Pricing and penetration assumptions come from “our internal expectations regarding pricing, adoption rates and service penetration across different geographic regions and economic environments.” The third-party sources provide the population base. The internal assumptions, applied to that base, produce a figure approximately twice what the same sources project for the entire global market.

The operational reality: 10.3 million Starlink Subscribers as of March 31, 2026, with ARPU of $66 per month, declining from $99 in 2023 [S-1, Connectivity segment data, p. 23]. Connectivity segment 2025 revenue of $11,387 million, with Q1 2026 annualized at approximately $13 billion. Current Starlink subscriber count is approximately 0.6% of the assumed 1.75 billion addressable household base. ARPU has declined 33% over two years as the customer base expanded into lower-priced markets.

The S-1’s own Risk Factors include the acknowledgment that “in our Connectivity segment, including Starlink broadband and Starlink Mobile, we face competition from terrestrial fixed network providers, mobile network operators, and other satellite providers, and our services may be less competitive in certain markets, including dense urban areas where terrestrial fiber and wireless networks may offer higher capacity, lower cost, or more consistent performance” [S-1, Risk Factors, p. 62]. The acknowledged competitive disadvantage applies to the highest household density and highest pricing markets.

The capital implication. Capital intensity for satellite broadband at scale is approximately 2.0x PP&E-to-revenue, consistent with terrestrial broadband operators (Comcast at 2.2x, Charter at 2.0x, Deutsche Telekom at 2.3x, BT Group at 2.5x), with satellite broadband comparables (HughesNet/Viasat at 1.5–1.7x, Eutelsat OneWeb at 3.3x in early stage), and with SpaceX’s own current Connectivity capital intensity of approximately 2x.

A 10% share of the $870 billion TAM by 2030 — approximately $87 billion in Broadband revenue, which from a 2025 base of approximately $10 billion implies a 50% compound annual growth rate consistent with recent trajectory — requires PP&E of approximately $174 billion at industry-standard intensity. Current Connectivity PP&E allocable to broadband is approximately $20–22 billion. Incremental PP&E required through 2030 for a 10% share: approximately $150–155 billion.

The Connectivity segment’s own EBITDA generation supports a portion of this capital deployment. 2025 segment Adjusted EBITDA was $7,168 million; growth trajectories consistent with reaching 10% TAM produce cumulative segment EBITDA through 2030 of approximately $80–120 billion. The Broadband-specific incremental capital requirement of $150–155 billion exceeds the segment’s full cumulative EBITDA generation; cross-segment cash flow or external financing supplies the remainder.

Connectivity Mobile — $740 Billion

The Connectivity Mobile construction follows the same methodology type — count of addressable devices multiplied by ARPU and monthly service revenue. The named sources include Omdia’s “Mobile Forecasts Summary” [S-1 citation xxiv], the Ericsson Mobility Report [S-1 citation v], J.D. Power’s wireless network quality analysis [S-1 citation xii], the CTIA’s spectrum shortfall report [S-1 citation ii], and the Global Satellite Operators Association’s “Satellite Solutions for Universal Service” [S-1 citation vii].

The reproducible math: approximately 8 billion mobile devices in active use globally; excluding China and Russia reduces the base by approximately 1.5 billion devices to roughly 6.5 billion addressable. At an assumed weighted-average ARPU of approximately $9.50 per month, 6.5 billion × $9.50 × 12 produces $741 billion, matching the $740 billion disclosed.

The triangulation point: Omdia’s 2030 global mobile service revenue projection is approximately $1.1–1.2 trillion for the entire global mobile services market — every mobile carrier, every device, every service tier combined. The satellite-to-mobile component within that projection is approximately $5–15 billion by 2030 in Omdia’s published scenarios. Ericsson’s adjacent projection has the satellite-to-mobile component at $10–20 billion. Broader analytical work from Counterpoint Research, Northern Sky Research, and Analysys Mason places the entire satellite-to-mobile market — AST SpaceMobile, Iridium, Globalstar, Starlink Mobile combined — in the $15–30 billion range by 2030.

The SpaceX-claimed $740 billion Mobile TAM is approximately 25–50x what the broader industry analytical apparatus projects for the entire satellite-to-mobile market in 2030.

The structural feature: Starlink Mobile is positioned in the S-1’s own Business section as “supplementing terrestrial networks and substantially reducing mobile ‘dead zones’” [S-1, Business — Our Leading Capabilities, p. 16]. It is positioned as the supplementary layer, not as primary mobile service. The TAM construction treats every mobile device globally as addressable at approximately $9.50 per month — every mobile device subscribing to supplementary satellite coverage at roughly 10–15% of typical wireless ARPU.

The operational reality: 7.4 million monthly unique devices across approximately 30 countries, served by approximately 650 V1 Mobile satellites [S-1, Prospectus Summary, p. 8]. Current Mobile revenue is bundled within the Connectivity segment and not separately disclosed; industry estimates place it in the $100–260 million annualized range through MNO partnership revenue sharing. Current device count is approximately 0.1% of the assumed 6.5 billion addressable base. The multiplier from current operations to TAM is approximately 3,000–7,000x.

The S-1’s own Risk Factors acknowledge the operational dependencies that the TAM construction does not address: “The expansion of our satellite-to-mobile connectivity services depends substantially on our ability to secure and maintain partnerships with mobile network operators and on the adoption of necessary hardware and software modifications by device manufacturers... To achieve full 5G NR-NTN compliance and optimal performance would likely require handset manufacturers to implement hardware and software modifications, primarily to the radio-frequency front end, in future devices. We do not have direct contractual arrangements with handset manufacturers; instead, we expect MNO partners, as major purchasers of mobile devices, to encourage or drive such adoption. There can be no assurance that these modifications will be adopted on our preferred timeline, or at all” [S-1, Risk Factors, p. 62].

The Mobile service depends on the EchoStar Spectrum Transaction ($11.5 billion plus 261.8 million shares of Class A common stock) closing in November 2027, on international spectrum approvals being granted country by country, on handset hardware modifications being adopted by device manufacturers, and on MNO partnerships in markets representing the addressable device base (currently 30 of approximately 200 countries). All four dependencies are present-tense constraints; the TAM is not.

The capital implication. Capital intensity for dedicated satellite-to-mobile infrastructure at scale runs at approximately 2.5–3.0x PP&E-to-revenue, somewhat higher than broadband because the V2 Mobile constellation requires dedicated satellites separate from broadband satellites. A 10% share of $740 billion — approximately $74 billion in Mobile revenue by 2030 — implies PP&E of approximately $185–222 billion at industry-standard intensity. Current Mobile-allocable PP&E is approximately $3–5 billion (650 V1 Mobile satellites plus pro-rata ground and manufacturing). Incremental PP&E required through 2030 for a 10% share: approximately $195–200 billion. Plus the EchoStar Spectrum Transaction at $11.5 billion in cash and 261.8 million Class A shares.

The Mobile capital requirement runs alongside, not within, the Broadband requirement. Combined Connectivity incremental PP&E for 10% share targets by 2030: approximately $345–355 billion.

AI Infrastructure — $2.4 Trillion

The AI Infrastructure construction follows the methodology described in the TAM Risk Factor paragraph: projections of global data center compute demand from third-party sources, combined with internal assumptions on the portion of global compute capacity utilized for AI workloads, power usage, utilization rates, and pricing. The named primary source is RAND Corporation’s “AI’s Power Requirements Under Exponential Growth,” January 28, 2025 [S-1 citation xxvi].

The reproducible math: RAND projects approximately 235 GW of global data center demand by 2030. Internal assumption of approximately 70% AI workloads produces approximately 164.5 GW of AI compute capacity. At hardware density of approximately 1.3 kW per GPU (H100 class accounting for cooling overhead at PUE 1.2), 164.5 GW divided by 1.3 kW equals approximately 126.5 million GPUs. At Silicon Data’s 2025 median H100 rental rate of $3.33 per hour [S-1 citation xxx], 80% utilization, 8,760 hours per year: $3.33 × 7,008 productive hours × 126.5 million GPUs = approximately $2.95 trillion. Within range of $2.4 trillion at modest input adjustments.

The triangulation point: McKinsey’s “The Cost of Compute: A $7 Trillion Race to Scale Data Centers” [S-1 citation xvi] projects $7 trillion in cumulative data center capex through 2030 globally — the supply-side construction cost, not the revenue. McKinsey’s 2030 AI services revenue projection in the same and adjacent reports is in the $1–2 trillion range. Bain & Company projects 2030 total AI market at $1.5–2 trillion combined services and infrastructure. Gartner projects 2030 AI infrastructure spending at approximately $400–500 billion annually. IDC projects 2030 AI infrastructure market at $500–700 billion. Goldman Sachs Research projects total AI market at $1–1.5 trillion by 2030.

The consensus range from major analytical firms is $1–2 trillion for the total AI market in 2030. The SpaceX-claimed $2.4 trillion AI Infrastructure TAM alone, before consumer subscriptions, digital advertising, or enterprise applications are added, is at or above the high end of total AI market projections from major analysts.

The pricing input deserves separate triangulation. The construction uses Silicon Data’s 2025 median H100 rental rate of $3.33 per hour. The Silicon Data report itself documents the rate declining from approximately $8 per hour in early 2023 to $3.33 by Q4 2025 — a 60% decline over approximately 24 months. Industry projections for AI compute rental rates through 2030 generally show continued decline as supply matures and as new chip generations (Blackwell, Vera Rubin) cycle in. At a 2030 projected rental rate of approximately $1.75 per hour, consistent with the cited source’s documented trajectory, the same construction with the same power and utilization assumptions produces approximately $1.24 trillion — roughly half the $2.4 trillion figure. The price held constant from 2025 is the input producing the difference.

The operational reality: SpaceX’s AI compute facilities, COLOSSUS and COLOSSUS II, collectively provide approximately 1.0 GW of nameplate compute draw [S-1, Business — AI Compute Infrastructure, p. 18]. Cumulative AI segment capex through Q1 2026: approximately $19 billion (2024: $5.6 billion; 2025: $12.7 billion; Q1 2026: $7.7 billion) [S-1, Capital Expenditures, p. 23]. Demonstrated capital intensity: approximately $19–20 billion per GW of deployed capacity. AI segment 2025 revenue: $3,201 million, with 2025 operating loss of $6,355 million [S-1, AI segment data, p. 23].

The Anthropic compute services agreement, executed in May 2026, provides $1.25 billion per month through May 2029 with reduced ramp in May and June 2026 [S-1, Prospectus Summary — Compute Services Agreements, p. 9]. At full ramp, approximately $15 billion annual revenue contribution. The S-1 notes the company expects “to enter into additional similar services contracts.”

The capital implication. Industry-standard greenfield AI data center buildout costs at scale are approximately $30–40 billion per GW, including land, power, structure, and compute hardware, with GPUs accounting for 60–70% of total cost per JLL’s “2026 Global Data Center Outlook” [S-1 citation xiv]. SpaceX’s demonstrated cost is approximately $19–20 billion per GW, reflecting both rapid deployment (122 days for COLOSSUS, 91 days for COLOSSUS II) and the structural advantage the S-1 claims: “data center construction costs for COLOSSUS II that are considerably lower than industry benchmarks on a per megawatt basis” [S-1, Business — Our AI Compute Infrastructure Advantage, p. 19]. The chip cost component remains exposed to market pricing for compute hardware.

A 5% share of global AI compute capacity by 2030 (approximately 8.2 GW of 164.5 GW global AI capacity) at midpoint capital intensity of approximately $27 billion per GW: approximately $220 billion in incremental capital. A 10% share at midpoint intensity: approximately $445 billion. Range across favorable and industry-standard assumptions: $160–578 billion. Midpoint scenario at approximately 7.5% share and midpoint intensity: approximately $335 billion in incremental capital through 2030.

The disclosed commitments include items that bear directly on this capital requirement. The Anthropic agreement is a revenue commitment that partially funds the capital deployment. The Cursor option, at $60 billion equity value with $10 billion in services and termination fees, reflects rather than reduces the underlying capital intensity — hyperscale AI participation requires capital at scale. Terafab, the chip manufacturing initiative with Tesla and Intel, is described in the S-1 as “subject to separate negotiations and agreements (including any development timelines, milestones and capital expenditures) [that] have not yet been determined” [S-1, Prospectus Summary, p. 4]. Industry estimates place cumulative Terafab capital at approximately $55 billion if it advances to full chip manufacturing scale. Each of these is evidence that participating in AI infrastructure at scale is more capital-intensive than industry-standard benchmarks, not less.

Plus orbital AI compute infrastructure on top. The S-1 describes deployment beginning as early as 2028 of “100 gigawatts per year via satellites carrying over 100 kilowatts of compute power per metric ton” requiring “thousands of launches per year and the transport of approximately one million metric tons to orbit annually” [S-1, Business — Orbital AI Compute, pp. 19–20]. Capital required for meaningful orbital AI compute capacity through 2030: additional $50–100 billion beyond terrestrial buildouts.

Consumer Subscriptions — $760 Billion

The Consumer Subscriptions construction is a population-times-ARPU build. The primary input source is Euromonitor International Passport 2026 Edition for population data [S-1 citation vi]. The YouGov report on AI usage and trust [S-1 citation xli] provides supporting data on consumer adoption rather than market size.

The reproducible math: approximately 5.5 billion people aged 10+ globally, excluding China and Russia. At an assumed weighted-average ARPU of approximately $11.50–12.00 per month, 5.5 billion × $11.75 × 12 produces approximately $776 billion, within rounding of the $760 billion disclosed. The pricing assumption is calibrated against existing consumer AI subscription tiers (SuperGrok at $30 per month, ChatGPT Plus at $20 per month, Gemini Advanced at $20 per month, Claude Pro at $20 per month), weighted down for free-tier conversion and emerging-market pricing.

The triangulation point: Counterpoint Research projects the 2030 consumer AI subscription market at approximately $40–60 billion. Bain projects $50–75 billion. Gartner projects the consumer AI software market at approximately $45–65 billion in 2030. The SpaceX-claimed $760 billion addressable is approximately 12–19x the broader analytical apparatus projects for the entire consumer AI subscription market.

The operational reality: Grok has approximately 117 million monthly active users using AI features as of March 31, 2026, of the 550 million total MAUs across Grok and X [S-1, Business — AI segment, p. 17]. Paid subscribers are a subset; industry estimates place xAI paid subscribers in the 2–5 million range. Current Grok consumer subscription revenue: approximately $200–500 million annualized. Multiplier from current operations to TAM: approximately 1,500–4,000x.

The capital implication. Consumer subscription businesses are software/platform businesses with PP&E-to-revenue at approximately 0.2–0.4x. The compute infrastructure supporting inference is captured in the AI Infrastructure capital base; the consumer-facing platform itself is light on incremental capital. For a 5–10% share by 2030 ($38–76 billion revenue), incremental PP&E beyond what is counted in AI Infrastructure: approximately $5–15 billion.

The Consumer Subscriptions line does not contribute materially to the cumulative capital base relative to the larger infrastructure-intensive components. Its analytical significance is what the multiplier reveals about the construction’s methodology — the same methodology family that produced the Connectivity Mobile $740 billion figure.

Digital Advertising — $600 Billion

The Digital Advertising component sources the $600 billion figure from S&P Global Market Intelligence's 2025 measurement of global digital advertising spending. The S-1 positions the opportunity as addressable for the X platform, integrated with Grok and the broader xAI ecosystem.

The structural feature: the global digital advertising market is concentrated in a small number of platforms. The top four digital advertising platforms (excluding China) hold approximately 80% of global digital advertising revenue. Google generated approximately $260 billion in advertising revenue in 2025; Meta approximately $160 billion; Amazon approximately $55 billion; ByteDance ex-China approximately $25–30 billion. Google and Meta combined hold approximately $420 billion of the global digital advertising market. The $600 billion SpaceX-claimed addressable is approximately 1.4x the combined current advertising revenue of Google and Meta.

X is currently in the second tier of digital advertising platforms, with revenue less than 1% of the leaders. The market leadership positions of Google, Meta, and Amazon are entrenched through decade-plus accumulation of audience, data, advertiser relationships, and advertising technology infrastructure. The TAM positioning of $600 billion as addressable for X — through integration with Grok and AI-mediated advertising — implies either substantial share displacement from the entrenched incumbents or a category redefinition where AI-mediated advertising is treated as additive to traditional digital advertising rather than displacing portions of it. The construction does not address what mechanism would produce either outcome at the scale required.

The capital implication. Digital advertising platforms are software and platform businesses with low PP&E intensity. For a 5–10% share by 2030 ($30–60 billion in revenue), incremental PP&E is approximately $6–24 billion beyond what is counted in AI Infrastructure.

Enterprise Applications — $22.7 Trillion

The Enterprise Applications component is the largest single line in the TAM disclosure — 80% of the total $28.5 trillion, 86% of the $26.5 trillion AI segment total. The construction derives from a single source: the Digital Cooperation Organization’s “Digital Economy Trends 2026,” December 2025 [S-1 citation iv].

The DCO is an intergovernmental organization founded in 2020 by Saudi Arabia, Bahrain, Jordan, Kuwait, Oman, and Pakistan, with subsequent member additions. Its 2026 Digital Economy Trends report measures the global digital economy at approximately $22.7 trillion. The DCO’s definition of the digital economy: economic activity that is “reliant on, significantly enhanced, or enabled by digital technologies.”

The methodology is a sector-spanning GDP-contribution measurement. The DCO measures the digital economy across approximately 12 categories: digital infrastructure (telecommunications, data centers, internet services); digital trade and e-commerce; digital financial services; digital health, education, and government services; software and IT services; AI and machine learning applications; data services and analytics; platform economies (social media, sharing economy, marketplaces); digital media and entertainment; IoT and connected devices; and digital manufacturing and automation. The aggregate measurement captures total economic activity across these sectors — Amazon’s e-commerce GMV, Walmart’s online sales, Netflix’s subscription revenue, Visa and Mastercard’s payment processing, the entire global telecommunications industry, the SaaS industry, online advertising, cloud computing, IT services, digital banking, online education, online healthcare.

Against 2026 global GDP per the World Bank of approximately $115 trillion [S-1 citation xxxviii], the DCO’s $22.7 trillion measurement is approximately 19.7% of global GDP.

The triangulation point: industry-standard measurements of the global enterprise applications and software market — software and services that businesses purchase to manage their operations — place the 2030 market at approximately $1.1–1.4 trillion. Gartner’s 2030 global enterprise software market: $1.1–1.3 trillion. IDC’s 2030 enterprise software market: $1.2–1.4 trillion. McKinsey’s 2030 AI-enabled enterprise software market: $300–500 billion. Bain’s 2030 enterprise AI services: $200–400 billion.

The SpaceX-claimed $22.7 trillion enterprise applications TAM is approximately 16–20x the conventional industry measurement of the entire enterprise software and applications market in 2030. The construction reaches an order-of-magnitude higher figure because it sources a fundamentally different measurement.

The operational reality: SpaceX has effectively no current revenue from enterprise applications as the term is conventionally used. Macrohard is described in the S-1 as currently in development [S-1, Prospectus Summary, p. 5]. xAI Gov and Grok Enterprise are early-stage offerings without disclosed revenue. The Anthropic compute services agreement is enterprise compute infrastructure, not enterprise applications. There is no current SpaceX enterprise applications business at meaningful scale. Against the $22.7 trillion claimed addressable, current revenue from this category is approximately zero.

The capital implication. Conventional enterprise applications businesses have low PP&E intensity — software and services with PP&E-to-revenue of approximately 0.1–0.3x. Compute infrastructure supporting AI-enabled applications is captured in the AI Infrastructure capital base. The enterprise applications business itself, separate from infrastructure, requires minimal additional PP&E.

But the relevant capital question is not PP&E. It is the operational architecture required to build and operate an enterprise applications business at any meaningful share of a $22.7 trillion claimed market. The S-1 does not allocate capital expenditure to enterprise applications as a category. The disclosed capex breakdown is by segment — Space, Connectivity, AI — with the AI segment capex covering compute infrastructure, model development, and platform development, not enterprise applications operations.

To address an enterprise applications market at even 0.1% share ($22.7 billion in revenue) requires a sales force in the tens of thousands of people; customer success and professional services organizations; a partner channel and systems integrator ecosystem; industry-specific go-to-market organizations; compliance, security, and procurement infrastructure for enterprise sales; and geographic expansion into every major market globally. None of these operational requirements is reflected in current SpaceX operations or disclosed commitments.

The Enterprise Applications line, taken on its own terms: methodology is single-source attribution to a GDP-adjacent measurement of the global digital economy; magnitude is 80% of the total TAM disclosure and approximately 20% of 2026 global GDP; no capital is allocated to it in the disclosed commitments; no operational architecture is described in the business operations; current revenue from this category is approximately zero.

And one further structural observation. SpaceX is operationally a space and connectivity company — current revenue concentrates in Connectivity; current capital expenditure concentrates in Space and AI infrastructure. The TAM disclosure positions the company differently. Eighty percent of the claimed addressable market is enterprise applications, a category in which the company has no current operations, no allocated capital, and no disclosed operational architecture. The TAM presents what the addressable market would require the company to be. The operations and capital structure reflect what the company actually is.

Caption: At industry-standard capital intensity applied to operationally defensible share targets for the six methodology-bearing components, the implied capital base through 2030 is approximately $855 billion. The IPO raise of $50–75 billion and the framework’s $235 billion aggregation of disclosed commitments through 2030 are shown for reference. The Enterprise Applications component, despite its 80% share of the TAM, has no allocated capital in the disclosed commitments.

Observed and Unobserved

The S-1 is a regulatory disclosure document. It observes substantial operational reality: current revenue and segment breakdown; subscriber counts and ARPU trajectory; AI compute capacity and demonstrated deployment; capital expenditure run rates by segment; the balance sheet including $54 billion in PP&E, $29 billion in outstanding debt, $41 billion in accumulated deficit. It observes substantial forward commitments: the Bridge Loan; the Anthropic compute services agreement; the Cursor option and termination/services fee structure; the Valor acquisition; the EchoStar spectrum transaction with its cash and share components; the Terafab framework agreement with Tesla and Intel; the ongoing investment requirements of the AI segment.

The TAM construction sits within this regulatory architecture but operates differently. The TAM is presented as opportunity assessment — what is addressable — without the corresponding disclosure of what addressing it would require, when, or how.

What the S-1 does not disclose, against the TAM construction:

— The anchor year for the TAM. The disclosure is present tense; the inputs span present-year measurements through 2030-projected demand to 2025-priced compute held constant. Industry analytical convention requires specified forecast horizons; the S-1’s TAM does not specify one.

— The capital base required to address the TAM. The S-1 discloses capital expenditure run rates through 2030 in separate sections but does not aggregate them and does not project the capital base required to address the claimed addressable market at scale. This analysis establishes that at industry-standard capital intensity for operationally defensible share targets in the six methodology-bearing components, the requirement is approximately $850–900 billion in incremental capital by 2030. The S-1 does not perform this calculation or disclose its result.

— The operational architecture for Enterprise Applications. The $22.7 trillion claim implies an operational base that the S-1 does not disclose. No enterprise applications segment is presented in the financial statements. No sales force, customer success organization, partner channel, or professional services capability is described in the business operations. No capital is specifically allocated to building such operations.

— The methodology for Digital Advertising. The six other components have identifiable source attribution from the industry data citations. Digital Advertising is the exception. The construction is not visible from the front-matter disclosures.

— The trajectory of compute hardware rental pricing through 2030. The S-1 cites Silicon Data’s H100 Rental Price Over Time analysis [citation xxx], which documents the rate declining from approximately $8 per hour in 2023 to $3.33 in 2025. The AI Infrastructure TAM construction uses the 2025 figure forward to 2030 without addressing the documented price decline trajectory.

— The share assumptions baked into each TAM component. The S-1 reports the addressable market; the implied capture assumption is not disclosed. This analysis adopts operationally defensible share targets (10% for Connectivity, 5–10% for AI Infrastructure) for the purpose of capital intensity calculation; the S-1 itself provides no such targets.

— The federal procurement environment the TAM assumes. The Risk Factors disclose the current 20% federal revenue concentration and acknowledge federal dependency as a risk factor [S-1, Risk Factors, p. 57]. The TAM construction depends, particularly for Space and indirectly for other segments, on federal procurement expanding at scale. The S-1 does not disclose the federal procurement environment its TAM assumes.

Each of these is a disclosure that the regulatory architecture does not require. The S-1 is compliant with disclosure standards as those standards are conventionally applied to TAM presentations. The framework reading is that the disclosure architecture permits the assertion of addressable market without the corresponding disclosure of what addressing it would require.

Cash Implications

The cash gap established in the prior analysis (”SpaceX, Adding It Up — The $235 Billion Cash Gap”) aggregated the disclosed cash commitments through 2030 — Bridge Loan, Anthropic compute services obligations, Cursor option mechanics, Valor acquisition, EchoStar spectrum, Terafab framework, and AI segment ongoing losses — at approximately $235 billion against an IPO raise of $50–75 billion. The aggregation was the framework’s analytical work; the S-1 does not present the $235 billion figure. The S-1 discloses the components in separate sections.

This analysis extends the framework’s work in a second direction. At industry-standard capital intensity applied to operationally defensible share targets for the six methodology-bearing TAM components, the incremental capital requirement through 2030 is approximately $850–900 billion. This figure is also the framework’s analytical work; the S-1 does not present it. The S-1 discloses the inputs — TAM by component, current operational data, demonstrated capital intensity in current operations — but does not perform the calculation.

The combined picture, from the framework’s analytical work applied to the S-1’s disclosed components:

— Disclosed forward commitments through 2030 (framework aggregation): approximately $235 billion.

— Implied additional capital for operationally defensible share targets in the six methodology-bearing components (framework calculation): approximately $615–665 billion beyond the disclosed commitments.

— Total implied capital base through 2030: approximately $850–900 billion.

— Disclosed funding sources (IPO raise, current cash on hand, expected operating cash generation from profitable segments through 2030): approximately $200–275 billion.

— Implied funding gap by 2030: approximately $575–700 billion. Against an IPO raise of $50–75 billion, the coverage ratio is approximately 7–13%.

The funding architecture disclosed in the S-1 deserves separate attention. The S-1 names the IPO as the primary cash event. The Bridge Loan and Credit Facility provide cash borrowing capacity. Beyond these, the disclosed major strategic commitments are denominated in stock rather than cash: the Cursor consideration would be “shares of our Class A common stock based on an implied equity value of Cursor of $60.0 billion” [S-1, Prospectus Summary — Collaboration with Cursor, p. 9]; the EchoStar spectrum transaction includes a 261.8 million share component; the directed share program for employees and designated individuals is part of the IPO mechanics; future acquisitions and partnerships are presumed to follow similar structures.

The architecture works for one set of commitments and not for another. Stock-denominated consideration funds acquisitions where the seller accepts stock; it funds partnerships where the counterparty accepts stock-denominated obligations. The physical infrastructure that the TAM construction implies — satellites, data centers, launch vehicles, chip facilities, ground stations — is paid for in cash, ultimately sourced from operating cash flow, debt issuance, or equity issuance converted to cash in the capital markets. At the implied scale of the capital base, the equity-issuance-to-cash channel would require capital markets to absorb dilution at a magnitude that itself becomes a constraint. The disclosed funding architecture, structured for direct stock-denominated commitments, addresses one set of capital requirements. The physical infrastructure requirements at TAM scale require a different funding architecture than the one disclosed.

Three sources of capital are available at the implied scale. Two appear in the S-1’s funding architecture. The third is required by the math but is not named in the funding architecture.

Sustained debt expansion is disclosed. SpaceX has $29.1 billion in current outstanding principal indebtedness [S-1, Risk Factors, p. 60]. The Credit Facility was amended in May 2026 to increase borrowing capacity and extend the maturity date. The 22-bank underwriting syndicate listed on the prospectus cover provides a deep capital markets relationship. The structural feature: the implied capital base requires sustained debt issuance at substantial scale through 2030.

Sustained equity issuance is disclosed. The dual-class share structure preserves voting control while permitting substantial Class A share issuance for acquisitions and capital raises. The directed share program, the Cursor stock consideration, the EchoStar share component, and the broader Class A share supply mechanics all support continued equity issuance. The structural feature: the implied capital base requires sustained Class A share dilution at substantial scale, with voting control preserved through the dual-class structure.

Federal procurement at sovereign scale is required by the math but not named in the funding architecture. The Space TAM requires federal procurement expansion at scale — Artemis program expansion, Space Force contracts at materially larger run rates, NSSL expansion. The Connectivity TAM, particularly for satellite-to-mobile, requires federal commitment to spectrum access and FCC support. The AI Infrastructure TAM, particularly for compute supporting national security applications including xAI Gov and Starshield, requires federal commitment to AI infrastructure procurement at scale. The S-1 discloses the current 20% federal revenue concentration as a baseline but does not project the federal commitment expansion that the TAM construction implies.

Against the implied $575–700 billion funding gap, even sustained debt expansion and equity issuance at aggressive scale cannot close the gap without federal commitment that scales with the implied capital base. The math points at the source the disclosure does not name.

Closing

The math points at the construction. The TAM disclosure synthesizes inputs of different vintages from a citation list of 41 sources, anchored to no specific year, and produces a $28.5 trillion figure that 80% derives from a measurement of the global digital economy rather than the conventional enterprise applications market. The aspiration gets translated into something different in the disclosure.

The math points at the capital. At industry-standard capital intensity applied to operationally defensible share targets for the six methodology-bearing components, the implied capital base through 2030 is approximately $850–900 billion. Against disclosed funding sources of approximately $200–275 billion, the funding gap is approximately $575–700 billion. The disclosed funding architecture is structured for acquisitions and partnerships, not for the physical infrastructure that the TAM construction implies.

The math points at the governance. The disclosed governance and control structure provides controller discretion on the timing and cost of investments, separate from their executional risk. The controller decides when and on what terms capital gets deployed. The market bears the executional outcome.

Yesterday’s piece, SpaceX, Adding It Up — The $235 Billion Cash Gap, worked the present tense. It aggregated the cash commitments the S-1 discloses through 2030 — Bridge Loan, Anthropic compute services, Cursor option, Valor acquisition, EchoStar spectrum, Terafab framework, AI segment losses — at approximately $235 billion against an IPO raise of $50–75 billion. The aggregation was the framework’s; the S-1 discloses the components in separate sections without aggregating them.

This piece worked the implied tense. What the aspiration alone requires, before any execution, applied to the six methodology-bearing TAM components at operationally defensible share targets, is approximately $850–900 billion in incremental capital by 2030. The aggregation is also the framework’s; the S-1 makes the TAM claim and discloses the operational data, but does not perform the calculation that connects the two.

Two pieces remain before the framework reading is complete.

The next piece works the governance and the constraints. The dual-class share structure, the controlled-company exemption, the forum selection clauses, the arbitration provisions, the Texas Business Court election, the 22-bank underwriting syndicate, and the federal procurement architecture all bear on the question this piece holds out: through what mechanisms is the implied capital base supposed to be financed, and on what assumptions about the demand environment that sustains it. The math establishes the requirement. The governance and constraint structures determine how the requirement gets addressed, by whom, and on whose schedule.

The piece after that works the synthesis. The framework reading across the body of work — the disclosure architecture’s structural features, the capital intensity at sovereign scale, the controller discretion separated from executional risk, the federal commitment required by the math but not named in the funding architecture — assembles into a hypothesis the math will either support or not. The hypothesis is held until the math is in.

The aspiration alone, before any material participation in the claimed market is achieved, is the issue. The math points. The disclosure architecture does not resolve.

Recent updates and news

Six things came into focus since our May 10 piece on the cash gap:

• The S-1 itself, filed May 20, with detailed segment reporting and forward commitment disclosures.

• The IPO raise range: $50–75 billion gross. Valuation rumor at $1.75 trillion or above.

• The Anthropic contract: $1.25 billion per month through May 2029, 90-day cancellable. $45 billion headline against capacity that does not yet exist.

• The Bridge Loan terms: $20 billion at SOFR plus 0.75–1.75%, effective rate 4.58% as of March 31, 2026, executed March 2 to retire xAI and X debt at 9.5–12.5%. Covenanted to be repaid within six months of IPO proceeds receipt. Cost of the swap in Q1 2026: $1.163 billion in cash prepayment penalties and a $1.526 billion accounting loss on extinguishment.

• The Cursor option mechanics: 30-day call window opening seven trading days after IPO completion or September 30, 2026, whichever is earlier. Exercise consideration $60 billion in Class A common stock per the disclosed terms. Termination consequence $1.5 billion termination fee plus $8.5 billion deferred services fee in cash.

• The Grimes County tax abatement filing, May 6: Terafab Phase 1 at $55 billion, full buildout to $119 billion. Filed fourteen days before the S-1 disclosed no contractual Terafab commitment.

The estimated cash requirements, item by item

Putting every commitment in the same currency — necessary cash through 2030:

Total: approximately $235 billion in known cash and equity deployment through 2030.

The raise: $50–75 billion headline gross. After fees and Bridge Loan retirement, net usable in the $30–50 billion range.

The gap is 3 to 5 times the raise on these items alone.

How this works

Each item above is disclosed. None are aggregated. Each appears in a different section of the document, structured in a different form.

The Bridge Loan repayment is in the debt footnote. Use of Proceeds, where investors read about what the IPO funds, omits it. Investors reading only Use of Proceeds will not know that $20 billion is contractually committed to debt repayment within six months of closing.

The Anthropic contract is presented as forward revenue. The capex required to deliver the capacity Anthropic has purchased is folded into AI segment capex aggregated at $12.7 billion for 2025 and $7.7 billion for Q1 2026. Current COLOSSUS plus COLOSSUS II capacity is roughly 1.0 gigawatt total — supporting approximately 510,000 GPUs at 80 percent utilization. The S-1 says COLOSSUS II is being expanded to train Grok 5. Fulfilling Anthropic requires hundreds of thousands of additional GPUs of capacity, financed by capex the IPO is supposed to fund, against a contract Anthropic can cancel at 90 days.

The Cursor option is presented as an upside acquisition opportunity. The economic substance is something different. Exercise consideration is $60 billion in Class A common stock — equity dilution, not cash. Termination consequence is $10 billion in cash, payable as a $1.5 billion termination fee plus an $8.5 billion deferred services fee. Both outcomes are material. Whether the company can pay the termination amount in cash is the binding constraint that determines which outcome occurs.

Working from the March 31, 2026 balance sheet: $15.85 billion in cash plus $7.82 billion in marketable securities, against a $20 billion Bridge Loan repayment due within six months of IPO proceeds receipt, plus an $11.5 billion Spectrum cash component closing November 2027, plus Valor lease payments at approximately $3.4 billion per year, against a Q1 2026 free cash flow run rate of negative $36 billion per year. The $10 billion termination fee, added on top, is not comfortably payable in cash.

Which means the exercise decision is effectively pre-determined by cash constraint, not strategic merit. The 30-day call window opens seven trading days after IPO closing, expiring no later than the end of October 2026. Exclusivity prevents Cursor from accepting competing bids during the window, so SpaceX faces no competitive pricing pressure. Exercising in stock uses the newly-IPO’d equity currency at the IPO’s establishing valuation; declining requires paying $10 billion in cash the balance sheet does not comfortably support. The architecture makes the stock exercise the rational outcome on cash-constraint grounds, regardless of how the Cursor acquisition itself is judged on strategic grounds. The option, structurally, is a forcing function for the first major deployment of the IPO equity, on a schedule that ensures the acquisition prices at the IPO’s establishing valuation before the market has time to test the valuation against operating performance.

The Valor sale-leaseback is filed under the technical accounting classification “failed sale-leaseback.” The substance is $20 billion of related-party debt financing from the CEO of a private equity firm who sits on the SpaceX board. The classification does the work of describing related-party debt in the language of an accounting technicality.

The Spectrum Transaction is presented as an intangible asset acquisition for $19.6 billion, of which $11.5 billion is cash (debt payoff plus loans to EchoStar’s trust through November 2028). The equity consideration of 261.8 million Class A shares at a fixed $42.40 per share post-split price was set in November 2025; at the rumored $1.75 trillion valuation, EchoStar receives a markup on the equity consideration on top of the cash. The capex to deploy the spectrum into a revenue-generating network is folded into Connectivity segment capex.

Terafab is a “collaboration” with Tesla. The S-1 states “neither Tesla nor Intel are obligated to remain a part of the project, and we may not enter into any such definitive agreements.” The same project was filed with Grimes County two weeks earlier at $55 billion Phase 1, scaling to $119 billion. Same project, same officer, same fourteen-day window. The tax abatement filing represents the maximum credible project budget because the abatement scales with budget. The S-1 represents only what is contractually committed because forward capex commitments scale negatively against cash available for disclosed growth uses. Both filings are correct within their incentives. The set is not coherent.

The xAI operating losses are disclosed historically by segment. The forward projection of when those losses close, and the capex required to build the revenue that would close them, is not separately stated.

What the S-1 says about cash requirements

In substance, the Risk Factors section says:

• Capital requirements are significant and ongoing.

• The company may not be able to raise sufficient additional capital.

• Cash flow has been and will likely remain negative.

• Funding depends on continued access to capital markets.

These are qualitative warnings. They are not quantified against the commitments above. Investors are told cash needs are large, funding is uncertain, and operations are cash-flow negative. They are not told that aggregate cash deployment required to fulfill the disclosed operational commitments is approximately $235 billion against a $50–75 billion raise.

The aggregate is in the risk section as a warning. The aggregate is not in the financial sections as a number.

The company also discloses additional unquantified exposures that sit on top of the table above:

• A $399 million litigation accrual at March 31, 2026, with the company stating reasonable possibility of additional material losses that are not currently estimable.

• The Vidstream patent verdict, $105 million in damages plus $67 million in prejudgment interest, on appeal at the Federal Circuit.

• The Pampena v. Musk partial judgment, entered April 3, 2026, against Mr. Musk in his personal capacity on Section 10(b) violations connected to his 2022 Twitter purchase statements. The company discloses this in its own management section as a risk to the company.

• The Grok image-generation lawsuit cluster, including the Mayor and City Council of Baltimore complaint filed March 24, 2026, seeking statutory penalties and injunctive relief — claims not quantified.

• The European Commission Digital Services Act fine of €120 million, challenged February 16, 2026 in the General Court of the European Union — challenge pending.

• The NAACP Clean Air Act challenge to COLOSSUS II turbines, filed April 14, 2026, with a preliminary injunction motion seeking to enjoin operation.

• Indemnification obligations to directors, officers, and contractual counterparties, which the S-1 states “may not be subject to maximum loss clauses” and for which the company “has not accrued a liability.”

These exposures are disclosed qualitatively. None are aggregated. None are quantified against the cash bridge. The S-1 states the company believes there is a reasonable possibility it may incur material losses that exceed its estimates. Translated back to cash, these are additional draws against the same constrained pool.

The gap that is disclosed

It is disclosed three ways simultaneously:

In the financial statements, as historical results and committed obligations, item by item, in different notes.

In the operational claims, as forward intentions whose capex implications are aggregated or unstated.

In the risk factors, as qualitative warnings that capital needs are large, funding is uncertain, and litigation exposures may exceed estimates.

Putting it all back in the same currency, the three combine: approximately $235 billion in known forward commitments, plus the company’s disclosed additional unquantified exposures, against a $50–75 billion gross raise, against $6–7 billion per year of operating cash flow, against –$36 billion per year of free cash flow run rate. The S-1 contains every input. The S-1 does not perform the calculation.

Where the cash might come from

One pathway is the Tesla acquisition, speculated by Bloomberg, ION Analytics, Walter Isaacson, and the prediction markets, and structurally available given Mr. Musk’s roles at both companies, Tesla’s cash position, and the cross-shareholder mechanics already in place. Even if completed at scale, a Tesla absorption closes a fraction of the gap — Tesla’s approximately $45 billion of cash plus incremental debt capacity addresses a portion of the deployment, not all of it. The S-1 does not address the pathway in any forward-looking section.

A federal commitment pathway would require disclosed CHIPS Act expectations for Terafab and disclosed government contract commitments at TAM scale. Neither is in the document. Customer A (the U.S. government) is 20.9 percent of 2025 consolidated revenue, declining as a percentage from 25.2 percent in 2023.

A subsequent equity issuance pathway is the implicit mechanism: the IPO establishes the currency, the currency funds the gap through additional offerings, strategic transactions priced in the equity, and continued related-party financing. The Cursor mechanism described above is the first scheduled instance of this pathway: convert IPO equity to a strategic acquisition before the market tests the establishing valuation, on a timeline the option mechanics determine. The architecture does not require the company to perform this conversion. The architecture does require the company to pay $10 billion in cash if it does not. Given the balance sheet, the conversion is the cash-conserving outcome.

A continued private capital pathway — Valor leases extending, additional related-party financing, additional sale-leaseback structures — would scale the related-party debt from $20 billion to materially larger figures and would compound the governance asymmetries the controlled-company structure already permits.

Each pathway closes a fraction. None alone closes the gap. The question the S-1 does not answer, and that no single forward-looking section is required to address: where this amount of cash, on the timeline the operational commitments require, can be raised.

This piece continues the analytical work in Adding It Up — The $165 Billion Cash Gap (May 10, 2026), which established the common-currency translation method and identified the eight S-1 disclosure questions this update revisits against the document the company filed on May 20.

A lot is moving very quickly. Four major capital deployments in three months: the xAI absorption in February, the Cursor option in April, the Anthropic compute commitment in May, the Terafab Phase 1 filing the same week. Each piece has its analysts. Bernstein on Terafab capacity. New Street Research on Anthropic revenue. Palihapitiya and Sacks on the Cursor option mechanics. Reuters’ S-1 review on the xAI cash consumption. The trees are well-mapped. The forest less-so.

In absolute terms, $165 billion is hard to absorb in normal proportion. The CHIPS Act, in total, was $52.7 billion. Saudi Aramco’s record IPO raised $29 billion. The cash gap is more than three times the first and roughly six times the second. In relative terms, it consumes the IPO’s net usable proceeds three times over. Either lens makes the gap material to an investor’s decision.

Multiple sources have signaled the Tesla acquisition for months: Bloomberg in January, ION Analytics modeling the share-exchange mechanic, two Nevada “merger sub” entities formed January 21, Walter Isaacson publicly expecting it, the Polymarket and Kalshi prediction markets pricing it. There are now a lot of reasons. The pieces are in place. The math of what’s already been announced makes the case.

The consolidated picture

Four announced deals, plus the existing SpaceX baseline, plus the IPO, looked at for the next 5 years. The headline numbers:

• Cumulative cash gap, 2025-2030 (Cursor in stock): $165 billion

• Cumulative cash gap, 2025-2030 (Cursor in cash): $225 billion

• Total stock issuance: $250-310 billion (xAI plus Cursor if stock)

• 2030 revenue from announced deals (best case): $59.5 billion

• 2030 operating income from announced deals (best case): $19.9 billion

The 2030 operating income figure stands out. SpaceX standalone reported approximately $8 billion in operating income in 2025 on $15-16 billion of revenue. Five years from now, with $200 billion of new capital deployed across four announced commitments, consolidated operating income reaches $19.9 billion in the best case. Roughly 2.5 times the existing baseline, requiring more than 25 times the capital to get there. And every operational assumption has to hold simultaneously: Anthropic ramping its compute spend with SpaceX to $5 billion annually, Cursor winning the AI coding category, Terafab achieving the 2030 first-wafer milestone (Intel Ohio went from 2025 target to 2027-2028 actual delivery), xAI’s losses moderating on schedule, Starlink continuing its current trajectory.

Where the cash could come from

Five categories exist beyond the IPO. Each has constraints. The constraints stack.

The US government category deserves a closer look. It differs in nature from the others, operating through procurement decisions and budget appropriations rather than market pricing, and its incremental contribution net of associated capex is relatively small. This is consistent with the broader pattern: the load-bearing relationships in this architecture run through private capital markets and through Tesla, not through Washington.

The aggregate brackets the cash gap. The arithmetic appears to work, but only if every assumption holds at the upper end at the same time, with no friction, over five years. That configuration has no historical precedent at this scale.

The Tesla acquisition is the only single mechanism large enough to close the gap with execution margin. The pieces are in place: Tesla holds approximately $45 billion in cash, has incremental debt capacity, would absorb passive index demand at deal pricing, and would trigger compensation realization at both companies. Together they resolve the cash question through a single transaction rather than requiring five competitive marketplaces to succeed simultaneously. The conventional framings for a Tesla merger have focused on Musk’s strategic preference, operational synergies, governance simplification. Those reasons exist. The math of the past three months adds another.

What we’ll look for in the S-1

The S-1 lands within the next two weeks, per SEC rules requiring publication 15 days before roadshow. The roadshow targets the week of June 8. The questions we’ll be looking for it to answer:

• Does it reconcile the announced commitments to a multi-year financing plan, with proceeds, timing, and pricing for each source?

• Does it disclose the bridge loan repayment in detail, including what the $20 billion existing facility consumes from the headline IPO proceeds?

• Does it quantify the capex schedule for Terafab, Anthropic compute infrastructure, and ongoing Starship development separately, so readers can see what the IPO actually funds?

• Does it address the Cursor option timing, exercised at IPO pricing in stock or held until the $10 billion partnership window closes?

• Does it acknowledge the cash flow impact of xAI’s 2025 operating loss continuing through 2026-2028?

• Does it discuss the Tesla relationship, beyond the existing $2 billion Tesla-into-xAI investment that overrode a Tesla shareholder vote in January, in a way that lets investors assess the cross-shareholder dynamics?

• Does it provide CHIPS Act and state incentive expectations for Terafab with sufficient detail to inform the project’s net cash trajectory?

• Does it disclose customer concentration in the Anthropic compute revenue, and the contractual structure that determines whether the $4-5 billion headline is realistic?

The risk-factor section will mention these. The question is whether the financial sections quantify them. Whatever the document discloses, the gap exists in absolute terms and matters in relative terms. The S-1 is the document where the company addresses the gap.

Showing our work

xAI absorption (February 2026)

Cursor option (April 2026)

Anthropic compute deal (May 2026)

Terafab Phase 1 (May 2026)