SpaceX's S-1 and first 10-Q: what the numbers say that the coverage does not

Short answer: SpaceX registered on Form S-1 on May 20, 2026, went effective on June 11, and sold 638,888,888 Class A shares at $135.00 on June 12 for net proceeds of $85,675 million. That much has been written about at length. What has been written about far less is the document that followed on August 4: the company's first Form 10-Q as a public registrant. Read together, the two filings describe a business whose capital allocation no longer matches its name, whose only profitable segment does not cover the losses of the other two, whose largest depreciation charges have not yet reached the income statement, and which paid for two acquisitions in the same currency on two opposite terms. This file works through what the numbers say, in the company's own figures, and draws out the parts that apply to companies a thousandth of its size.

1 · What the filings establish

Everything in this section comes from the company's EDGAR file, read on August 25, 2026.

Space Exploration Technologies Corp., a Texas corporation, CIK 0001181412, filed a registration statement on Form S-1 on May 20, 2026 under file number 333-296070. It was amended on June 3 and declared effective on June 11 at 10:00 a.m. The Class A common stock was registered under Section 12(b) and listed on Nasdaq under the symbol SPCX. The final prospectus was filed under Rule 424(b)(4) on June 12. The offering sold 638,888,888 shares at $135.00 per share including the full over-allotment exercise, producing net proceeds of $85,675 million after $575 million of underwriting commissions and offering costs. Ten book-running managers and twelve co-managers were named.

Two structural events preceded it. In May 2026 the board effected a five for one forward stock split, and all share figures in the filings are retroactively adjusted for it. On February 2, 2026 the company completed its acquisition of X.AI Holdings Corp., which had itself acquired X Holdings Corp. in March 2025. The 10-Q describes both as common control mergers. That single phrase carries more weight than most readers give it, and section 6 returns to it.

Then came the first quarterly report, filed August 4 for the quarter ended June 30, and a Form 8-K on August 14 reporting that the Cursor merger had closed. The disclosure controls conclusion was effective. There was no going concern paragraph, no material weakness, no late filing. On the compliance surface, this is a clean start.

2 · The company is now three companies, and only one of them earns

The 10-Q reports three operating and reportable segments under ASC 280, with the chief executive identified as the chief operating decision maker: Space, Connectivity and AI. The segment note is the most informative page in the filing, and it is worth reading before anything else.

All figures in $ millions as filed, except ratios, which are computed. Source: Form 10-Q, Note 18.

Connectivity, which is Starlink, produced $7,548 million of revenue and $2,844 million of operating income at a 37.7% margin. It is a large, profitable, growing subscription business. Starlink subscribers doubled year on year, from 6.0 million to 12.0 million. On its own it would be a serious public company.

Space, the rocket business the company is named for, produced $1,581 million of revenue and lost $1,204 million. Its revenue fell slightly against the prior year, from $1,611 million, because the segment only recognises revenue on customer launches, and customer launches fell from 21 to 17 in the half. The launch cadence did not collapse; the internal share of it rose. Of 77 Falcon launches in the half, 60 were internal.

AI produced $3,379 million of revenue and lost $3,726 million. Its operating margin is negative 110.3%, meaning the segment spends more than two dollars for every dollar it brings in.

Now put the three together, which the filing does not do for the reader.

Connectivity's entire operating profit does not cover the losses of the other two segments. It covers 58% of them. The remaining 42%, and all of the interest expense, comes from somewhere other than the operating business. This is not a criticism. It is a description of a company in a deliberate investment phase, and the company has been open about that intention. But it reframes the question a reader should ask. The question is not whether Starlink is a good business, because the filing shows that it is. The question is how long the gap is meant to run, and what closes it.

3 · Capital allocation is the real strategy statement

Companies describe their strategy in the business section. They reveal it in the capital expenditure line. Here the two are not saying quite the same thing, and the segment note gives the reader the evidence.

In the first half of 2025, AI took 47.6% of capital spending. In the first half of 2026, it took 82.7%. Space, the segment that gives the company its name and its public identity, took 7.8%. Total capital spending rose 309% in twelve months, from $6,965 million to $28,476 million, and almost the whole of that increase went to one place.

Read as a ratio, the picture is starker still. Connectivity spends $0.36 of capital for every dollar of revenue it produces. Space spends $1.41. AI spends $6.97. The property, plant and equipment note shows what the money bought: servers and networking equipment at $34,771 million gross, up from $22,694 million six months earlier, and data centre infrastructure at $3,991 million. Satellites, by comparison, stand at $13,788 million and flight vehicle hardware at $1,557 million.

On the balance sheet, by gross carrying value, the largest single asset class at this rocket company is servers.

The cash flow statement completes the picture. Operations produced $3,466 million in the half. Capital spending consumed $28,476 million. Free cash flow, defined simply as the difference, was negative $25,010 million. The IPO and a $25,000 million bond issue covered it, along with the rest of the financing inflow of $100,291 million.

At the half year pace, capital spending annualises near $57 billion. The $110,675 million raised from the IPO and the notes together, held against that pace, is roughly two years of funding. That is a calculation from filed figures, not a forecast, and the company retains explicit flexibility: the liquidity section states that if data centre needs ramp more slowly, it may reduce capital expenditure in that segment and reallocate. Management states it believes it has sufficient funding for at least the next twelve months, which is the standard ASC 205-40 look forward.

4 · The depreciation has not arrived yet

This is the part that a reader trained on income statements will miss, because it is not on the income statement yet.

Depreciation and amortisation for the half was $5,290 million, up 78% from $2,970 million. In the AI segment alone it was $3,378 million, up 113%. Those increases already reflect assets that entered service earlier. They do not yet reflect most of what was bought this year.

Construction in progress stands at $12,554 million, 15.1% of gross property, plant and equipment, and by definition is not being depreciated. Beyond that, the $23,551 million of AI capital spending in this half is largely at the front of its depreciation life. The filing does not disclose the useful lives assigned to AI infrastructure, so the exact future charge cannot be computed from the public record. What can be said is directional: at a five to six year straight line life, that half year of AI spending alone would eventually carry something in the region of $3.9 billion to $4.7 billion of annual depreciation. The AI segment's operating loss for the half was $3,726 million.

This is arithmetic offered as illustration, not as a prediction of results, and the actual charge will depend on lives, timing of placement in service, and any impairment or change in estimate. But the direction is not in doubt. The cost of this year's building programme reaches the income statement over the next several years, and it reaches it whether or not the revenue does.

A related detail sits in the same note. Interest is capitalised during construction on significant long term projects, including AI data centres. Capitalised interest reduces reported interest expense today and raises depreciation later. It is entirely proper under ASC 835-20. It also means the reported interest expense of $1,293 million understates the economic cost of carrying the debt during the build.

5 · Two acquisitions, one currency, two opposite bargains

This is the section that has no equivalent in the commentary, and it is the one small company directors should read twice.

Within roughly nine months, SpaceX agreed two large acquisitions paid for in its own Class A shares. It structured them in opposite ways.

The Cursor merger was priced on a fixed value. The consideration was set at an implied equity value of $60.0 billion for Anysphere, Inc., with the number of shares determined by dividing that value by the volume weighted average closing price over the seven trading days before closing. The merger closed on August 14, 2026 and the 8-K reports the outcome: 389,289,254 Class A shares for the outstanding stock, 1,752,426 shares for vested restricted stock units, plus approximately 29,128,326 assumed restricted stock units and approximately 44,365,047 assumed options. Dividing $60.0 billion by 389,289,254 shares gives an implied price near $154 per share.

Under that structure the seller's value is locked at signing. If the buyer's share price falls between signing and closing, the buyer issues more shares to deliver the same value. The buyer carries the risk, in dilution.

The EchoStar spectrum purchase was priced the other way. The agreement, signed in September 2025 and amended in November 2025, sets total consideration of approximately $19.6 billion, of which the equity component is described in the filing as "approximately $11.1 billion in equity, payable through the issuance of approximately 261.8 million shares of the Company's Class A common stock at a fixed value of $42.40 per share."

The share count is fixed. The $42.40 reference reflects the company's valuation when the deal was negotiated, before the IPO. At the August 25, 2026 closing price of $137.95, those same 261.8 million shares would be worth roughly $36.1 billion. The equity transfer has not happened yet. The licences moved to a Nevada business trust on May 22, 2026 after FCC approval on May 12, and the shares are due at the Spectrum Acquisition Closing, which the company expects around November 30, 2027.

Two consequences follow, and both are hedged deliberately, because the outcome depends on a share price on a future date that nobody knows.

First, in economic terms, the seller captured the re-rating. EchoStar negotiated a fixed number of shares against a pre-IPO valuation, and the public market has since priced those shares materially higher. Whatever the accounting says, the value leaving the buyer is not $11.1 billion at today's prices.

Second, in accounting terms, ASC 805-30-30-7 measures equity issued as consideration in a business combination at fair value on the acquisition date. If the spectrum purchase is accounted for as an asset acquisition rather than a business combination, the measurement basis differs, and the company has stated that payments under the arrangement are being recognised as prepaid assets and will be recognised as intangible assets on closing. Either way, the recorded cost is likely to be measured off a share price at closing rather than off $42.40, which means the intangible asset that eventually lands on the balance sheet may be several times the stated equity figure. The filing does not quantify this, and no reader should assume a number for a date more than a year away.

One further feature of that transaction deserves a note of its own. The company agreed to make payments to the trust, structured as loans, to service EchoStar's debt. The filing states plainly that "there is no expectation of repayment as the loan payments are forgiven," and that they are accounted for as additional consideration. Expected payments are $1,241 million in 2026, of which $856 million had been paid at June 30, and $828 million in 2027, with a further $827 million possible if closing slips to November 2028. A loan that everyone agrees will be forgiven is consideration, and the company has accounted for it as consideration. That is the right answer, and it is the answer a great many smaller companies get wrong.

6 · What common control accounting did to the comparatives

The 10-Q describes the xAI merger and the earlier X merger as common control mergers. Under ASC 805-50, a transfer of businesses between entities under common control is not a purchase. The receiving entity records the assets and liabilities at their historical carrying amounts, no goodwill arises from the transaction, and the financial statements are presented as if the combination had occurred from the beginning of the earliest period in which the entities were under common control.

That is why an AI segment with $1,465 million of revenue appears in the six months ended June 30, 2025 comparatives of a company that did not legally own xAI until February 2, 2026. The comparatives were recast, correctly and in accordance with the standard.

The practical point for a reader is this. When the AI segment shows revenue growing from $1,465 million to $3,379 million, that is a real operating comparison of the same underlying businesses. But those businesses were assembled under one owner's control across several transactions, and the accounting for bringing them together produced no step up in asset values and no acquisition goodwill. Goodwill on the balance sheet actually fell over the half, from $11,809 million to $11,645 million. A reader who assumes the AI business was bought and fair valued in the ordinary way would be reading the balance sheet wrongly.

It also means the assets supporting a segment that lost $3,726 million in six months sit at historical carrying value rather than at a value the market would set today. That cuts both ways, and it is a fact to hold, not a conclusion to draw.

7 · The related party balance sheet

Debt and finance leases on the balance sheet at June 30, 2026 carried at $39,364 million, being $2,525 million current and $36,839 million non-current. Of that, $13,329 million was owed to entities affiliated with Valor Equity Partners under equipment lease agreements for AI infrastructure hardware. The founder, chief executive and chief investment officer of Valor serves on the company's board.

A note on the denominator, because it matters. The MD&A states that the company and its subsidiaries had "$38,433 million in aggregate principal amount of indebtedness" outstanding. That is a principal figure. The balance sheet carries $39,364 million, which includes finance leases and reflects carrying value. The two are different measures and dividing one into the other produces a ratio that means nothing. Every percentage in this section is computed on the balance sheet carrying value, one basis throughout.

The note describes the April 2026 transaction as a failed sale-leaseback. That term is technical and it matters. When a company sells an asset and leases it back, ASC 842-40 sends you first to ASC 606 to ask whether control of the asset actually transferred to the buyer. If it did not, there is no sale. The asset stays on the seller's balance sheet, and the cash received is accounted for as a financing. That is precisely what happened here, and the filing says so.

The scale is what makes it notable. Related party lease financing grew from $4,507 million at December 31, 2025 to $13,329 million six months later, an increase of 196%. As a share of debt and finance leases on the balance sheet it went from 19.7% to 33.9% in six months. Interest expense to the related party was $513 million of $1,293 million total for the half, or 39.7%.

None of this is improper, all of it is disclosed, and related party financing is common where specialised equipment needs funding quickly. The observation is narrower: a reader who models this company's leverage without reading Note 17 will attribute a third of its debt and two fifths of its interest cost to arm's length lenders, and will be wrong.

Other related party dealing is smaller and also disclosed. The company purchased $295 million and $329 million of Megapack products from Tesla, Inc. in the three and six month periods, recorded in property, plant and equipment.

8 · Two customers, thirty eight percent of revenue, ninety days of notice

The revenue note discloses that Customer A represented 18.3% of consolidated revenue in the quarter and Customer B represented 19.5%. Together, 37.8% from two customers. Customer A spans all three segments. Customer B sits in the AI segment, and a year earlier did not reach the 10% disclosure threshold at all.

Applied to quarterly revenue of $7,814 million, that is roughly $1,430 million from Customer A and $1,524 million from Customer B in a single quarter. The AI Solutions and Infrastructure line grew from $311 million to $2,194 million year on year, a 605% increase, and the implied Customer B figure is close to 69% of that line.

The company added a new risk factor in this quarter to address it, and the risk factor is unusually specific about the contract terms: "Our cloud services agreements generally provide for monthly fees and, after an initial period inclusive of capacity ramp, may be terminated by either party upon 90 days' notice."

Set that against the capital commitment. The company is spending $23,551 million a half year on infrastructure whose largest revenue stream rests on monthly fee contracts cancellable on ninety days' notice. The asset life is measured in years. The contract life is measured in months. That mismatch is disclosed clearly and it is the single most important sentence in the risk factors.

Against that, the backlog note offers the other side. Backlog stood at $47,461 million, about 1.9 times annualised revenue, with 56% expected to be recognised within a year. Deferred revenue was $14,286 million, up from $12,116 million, meaning customers are pre-funding the business by a material amount. The Space and Connectivity enterprise contracts that make up most of that backlog look nothing like the ninety day cloud arrangements.

9 · What the tax line says that the price does not

The income tax note contains the sentence that a forensic reader stops on. As of June 30, 2026, the company "continues to maintain a full valuation allowance against its deferred tax assets in the United States," with narrow exceptions for certain state deferred tax assets and transferable investment tax credits.

A valuation allowance under ASC 740-10-30-5(e) is recorded when it is more likely than not that some or all of a deferred tax asset will not be realised. Realisation depends on future taxable income. So the company's own accounting judgment, formed with its auditors and disclosed in a certified filing, is that it is more likely than not that it will not generate enough United States taxable income in the relevant period to use its accumulated tax attributes.

That judgment sits in the same document as an accumulated deficit of $41,852 million and, in the market, alongside a capitalisation near $1.87 trillion and a published analyst consensus price target above the current price. The two are not contradictory. Valuation allowance assessment weighs objectively verifiable evidence, and a recent history of losses is the heaviest negative evidence there is; market prices weigh expectations. But the contrast is worth stating plainly, because the valuation allowance is management's own view, recorded under a standard that constrains optimism, and it is the more conservative of the two signals in the public record.

The effective tax rate was negative 0.6% for the half, against a 21.0% federal statutory rate, which the note attributes to jurisdictional mix and the valuation allowances.

One more line in the same neighbourhood. The company holds 18,712 units of Bitcoin with a cost basis of $661 million. Fair value fell from $1,637 million at December 31, 2025 to $1,098 million at June 30, 2026. Under ASU 2023-08 that $539 million decline runs through net income. It is roughly 11% of the $4,817 million net loss for the half. A rocket, broadband and AI company's reported earnings now move with the price of Bitcoin, and a reader reconciling the loss should know which part of it came from operations and which from a treasury position.

10 · Control

At June 30, 2026 there were 7,607 million Class A shares carrying one vote each and 5,569 million Class B shares carrying ten votes each. Class C, which carried no votes, was reclassified into Class A at the IPO, and Class D was eliminated. On those figures Class B holds 42.3% of the shares and 88.0% of the votes.

Dual class structures are common and legal, and Nasdaq permits them. The disclosure is complete. The number is offered because it is the answer to a question every governance-minded reader asks and few filings compute for you: what fraction of the votes accompanies what fraction of the ownership.

The Item 5 disclosure adds a related detail worth noting for its discipline. The chief financial officer and the president and chief operating officer both agreed to subject the vast majority of their shares to an extended lock-up, and both adopted Rule 10b5-1 arrangements that do not commence sales until 2027. A separate 10b5-1 arrangement was adopted by entities affiliated with a director for the potential distribution of up to 225,857,490 shares to their limited and general partners, expiring September 30, 2027. Under the 2023 amendments to Rule 10b5-1 these arrangements carry cooling-off periods and mandatory disclosure, and the company has disclosed them in the quarter of adoption, which is what the rule requires.

11 · Eleven things a small-cap issuer should take from this filing

Almost nothing above depends on scale. The mechanics that make this filing interesting are the same mechanics that decide whether a company with $30 million of revenue survives its next transaction. Set out plainly:

One. Your segment note is a strategy document, and you are one of its readers. Under ASC 280 segments follow how the chief operating decision maker actually allocates resources. Aggregating dissimilar operations to keep the disclosure simple does not only hide the picture from investors; it removes the one report that would have told the board where the money is going. SpaceX's segment note is the reason a reader can see that 82.7% of capital went to one place. Most small companies that aggregate cannot answer that question about themselves.

Two. Capital expenditure per dollar of revenue is a better strategy metric than any margin. $0.36 for Connectivity, $6.97 for AI. Run it by segment or by product line every quarter. It shows the shift long before the income statement does.

Three. Decide who carries the price risk before you agree a share deal, and say it in one sentence. Fixed value protects the seller and dilutes the buyer if the price falls. Fixed shares protects the buyer's share count and hands the seller the whole re-rating if the price rises. Neither is right in the abstract. What is always wrong is signing without knowing which one you agreed to. The gap between $11.1 billion stated and roughly $36 billion at today's price, on a single transaction, is the most expensive illustration of this point available in public filings.

Four. Collars, caps and walk-away rights exist for exactly this reason. If your deal is priced in your own shares and will take a year or more to close, the negotiation over price protection is worth more than the negotiation over the headline number.

Five. Depreciation follows capital spending, and the lag is where surprises live. Build the schedule for assets not yet in service before you commit to the spending, and show the board the year in which the charge peaks. Construction in progress at 15.1% of gross assets is a number every capital-intensive company should be able to state on demand.

Six. Test every sale-leaseback against ASC 606 before you book the gain. If control has not transferred, there is no sale, the asset stays on your balance sheet and the cash is debt. Getting this wrong is a common restatement source in smaller companies, because the gain on sale is usually the reason the transaction was proposed.

Seven. Related party financing needs to be disclosed as financing, tracked as leverage, and approved by people who are not on both sides. Item 404 of Regulation S-K sets a $120,000 threshold for disclosure. The governance question, which is separate, is who reviewed the terms. A third of debt and finance leases owed to a board-affiliated party is disclosed here in the ordinary course. In a small company with one or two independent directors, the same fact pattern with no independent review is how enforcement files begin.

Eight. Match your asset life against your contract life and disclose the mismatch yourself. Multi-year assets funding revenue on ninety day cancellable terms is a genuine risk, and the company wrote it into its risk factors in specific language rather than boilerplate. Specific risk factors are the ones that protect you. Generic ones are the ones plaintiffs quote back.

Nine. Your valuation allowance is a statement about your own future, made under oath. If you are telling investors the business is about to turn while your tax footnote maintains a full valuation allowance on the basis that future taxable income is not more likely than not, one of those two documents is wrong, and the audit committee should know which.

Ten. A forgiven loan is consideration. If you agree to make payments there is no expectation of getting back, that is part of the purchase price, whatever the paperwork calls it. Account for it that way from day one rather than after the auditors find it.

Eleven. Common control transactions do not create goodwill, and they do recast your comparatives. If you are reorganising entities under one owner before a raise or a listing, understand under ASC 805-50 what your prior-year figures will look like afterwards, and prepare the explanation before an investor asks why last year changed.

12 · Where it stands

As of August 25, 2026, SPCX closed at $137.95, above its $135.00 offer price, with a market capitalisation reported at approximately $1.87 trillion and a range since listing of $104.83 to $225.64. The Cursor merger closed on August 14. The Mesh Optical merger closed on July 6. The spectrum acquisition closing remains ahead, expected around November 30, 2027. The next 10-Q, for the quarter ending September 30, will be the first to carry Cursor's purchase price allocation, and the first to show a full quarter of the AI infrastructure that was under construction at the half year.

Those two disclosures, the purchase price allocation and the depreciation on newly placed assets, will say more about this company's economics than any launch. This page will not be updated for price movements; the figures above are stated as of their dates.

FAQ

What did SpaceX's S-1 and IPO actually raise?

The Form S-1, file number 333-296070, was filed May 20, 2026, amended June 3, and declared effective June 11, 2026. The company sold 638,888,888 Class A shares at $135.00, including full exercise of the over-allotment, for net proceeds of $85,675 million after $575 million of underwriting commissions and offering costs. Stated use of proceeds is AI compute infrastructure, launch infrastructure and vehicles, satellite constellation capacity, and general corporate purposes.

Which SpaceX segment makes money?

Connectivity, the Starlink business. In the six months ended June 30, 2026 it produced $7,548 million of revenue and $2,844 million of operating income, a 37.7% margin. Space lost $1,204 million and AI lost $3,726 million over the same period, so the consolidated result was an operating loss of $2,086 million.

How much of SpaceX's capital spending goes to AI?

82.7% in the first half of 2026, or $23,551 million of $28,476 million total. A year earlier the AI share was 47.6%. Space received 7.8% in the most recent half.

Why does the fixed share price in the EchoStar spectrum deal matter?

The equity consideration is a fixed 261.8 million shares at a stated $42.40 per share, or approximately $11.1 billion, agreed before the IPO. Because the share count is fixed rather than the value, the seller receives the same number of shares regardless of the market price at closing. At the August 25, 2026 price of $137.95 those shares would be worth roughly $36.1 billion. The shares are due at the Spectrum Acquisition Closing, expected around November 30, 2027, and the measurement for accounting purposes will use the price at that date, which is not knowable now.

What is a failed sale-leaseback?

A sale and leaseback in which control of the asset does not transfer to the buyer under ASC 606, so no sale is recognised. The asset remains on the seller's balance sheet and the proceeds are accounted for as a financing rather than as a disposal. SpaceX discloses its AI hardware lease arrangements with Valor Equity Partners on this basis, carrying $13,329 million within debt at June 30, 2026.

What does a full valuation allowance mean?

Under ASC 740, a valuation allowance reduces a deferred tax asset to the amount more likely than not to be realised. A full allowance in the United States means management concluded it is more likely than not that the company will not generate sufficient United States taxable income to use those attributes. SpaceX maintained a full United States valuation allowance at June 30, 2026, with limited exceptions for certain state deferred tax assets and transferable investment tax credits.

How much voting control do the Class B shares carry?

Class A carries one vote per share and Class B carries ten. At June 30, 2026 there were 7,607 million Class A and 5,569 million Class B shares, which works out to 42.3% of the shares and 88.0% of the votes for Class B. Class C had no voting rights and was reclassified into Class A at the IPO.