Space Exploration Technologies Corp. (SPCX), Financial Analysis

Insights derived from analysing SEC filings. Independent analysis of Space Exploration Technologies Corp. Class A common stock, not a publication of the SEC.

This report is based on every filing SpaceX has made under CIK 1181412: one 424B4 prospectus, one Form 10-Q, eight Forms 8-K, three Forms S-1 and S-1/A, four Schedules 13G, ten Forms 3, one Form 4 and twenty one Form D notices. Accession numbers are in the Sources appendix, and market prices are closing prices of 3 September 2026. SpaceX listed on 12 June 2026, so there is no Form 10-K, the annual report, and no proxy statement for an annual shareholder meeting, yet.

Summary

In the quarter to 30 June 2026 the Connectivity segment, which sells Starlink satellite broadband, produced 54.9% of revenue, the AI segment, which builds AI models and runs the X platform, produced 32.8%, and the Space segment, which launches rockets and builds spacecraft, produced 12.3%. Launching rockets and building spacecraft produced 21.9% of the $18,674m of revenue the company reported for the year to 31 December 2025, and Space segment revenue for the first half of 2026 fell 1.9% year on year. The AI segment first appeared as a reportable unit in the quarter to 31 March 2026. Of the 77 Falcon launches in the first half of 2026, 17 carried a customer payload and 60 carried Starlink satellites.

The AI segment arrived by acquisition. xAI was acquired effective 2 February 2026 and X Holdings by xAI effective 28 March 2025, both between entities under common control, so the audited statements were retrospectively recast to include them. Every prior period figure in this report therefore covers the three combined entities.

The company listed on 12 June 2026. It has never filed an annual report, has never filed a proxy statement and has never held an annual meeting. Three audited years and two audited balance sheet dates exist, all of them inside the 424B4 prospectus of 12 June 2026 (0001628280-26-042639), alongside one quarterly report for the quarter ended 30 June 2026 (0001628280-26-052535). Five year series, prior year risk factor comparisons and multiyear guidance records cannot be built. Where the material does not exist this report says so and puts no estimate in its place.

Two financings three weeks apart reset the balance sheet. The offering sold 638,888,888 Class A shares at $135.00 with the option exercised in full, for $85,675m of net proceeds, and on 26 June the company issued $25,000m of senior unsecured notes across five tranches at a 5.855% weighted average coupon. Cash and marketable securities stood at $100,009m at 30 June 2026 against $39,364m of debt and finance leases, a net cash position of $60,645m. Capital expenditure of $28,476m in the first half exceeded revenue of $12,508m, and 82.7% of it went to the AI segment.

Elon Musk holds about 46% of the economics and about 85% of the vote through a 10:1 dual class structure with no time based and no ownership based sunset. Class B converts to Class A on almost any transfer, so his voting share rises whenever another holder sells. New investors supplied roughly half the capital ever paid into the company, bought 4.2% of the shares and hold about an eighth of the vote.

The Connectivity segment earned a 37.7% operating margin in the first half of 2026, higher than any listed satellite operator in the comparator set. The Space and AI segments lost $4,930m between them over the same half. No peer group answers whether the Connectivity segment funds the other two long enough for either to become profitable. At the 3 September 2026 close of $149.74 the market values the company at $1,973.84bn, 76.5 times annualised revenue, an order of magnitude above every profitable operator in the comparator set.

1 Revenue & Business Model

In the year to 31 December 2025, launch services and government spacecraft work produced $4,086m, or 21.9% of $18,674m of total revenue. Starlink, the satellite broadband service, produced $11,387m, or 61.0%. By the quarter to 30 June 2026 the mix had moved again: the Space segment 12.3%, the Connectivity segment 54.9%, and an AI segment, first reported as a separate unit in the quarter to 31 March 2026, at 32.8%. Of the 165 Falcon launches in 2025, 43 carried a customer payload and 122 carried Starlink satellites.

The company listed on 12 June 2026 at $135.00 per Class A share and has filed no annual report. Audited financial statements exist only inside the 424B4 prospectus (0001628280-26-042639) and cover 2023, 2024 and 2025. One 10-Q exists (0001628280-26-052535), for the quarter ended 30 June 2026. Every series in this chapter therefore runs three years, and the interim columns are unaudited.

The audited years also cover two acquired businesses. The prospectus presents a retrospective combination of entities under common control: xAI, acquired effective 2 February 2026, and X Holdings, acquired by xAI effective 28 March 2025. The AI segment's 2023 revenue of $2,961m is almost entirely the advertising business of the platform now called X. PricewaterhouseCoopers LLP audited the combined statements.

What the three segments sell

The Space segment, the part of the business that designs, manufactures and launches reusable rockets, books revenue two ways. Launch Services is point in time revenue from deploying customer payloads on Falcon 9 and Falcon Heavy, on fixed price contracts of one to five years. Launch & Development is over time revenue from spacecraft development and launch and mission services for government agency space programmes, on fixed price contracts running up to fourteen years. Starlink deployments generate no intersegment revenue; those launch costs are capitalised into satellites. Space revenue counts customer activity only, which is why cadence and revenue move apart.

The Connectivity segment, which operates the Starlink satellite broadband network, sells two things. Consumer is a monthly subscription priced by market and download speed, plus a one off kit sale. Enterprise & Government covers aviation, maritime, land mobility, fixed sites, Starshield contracts for United States government and national security customers, and revenue share with mobile network operators for Starlink Mobile. Consumer subscribers produced over 60% of Connectivity revenue in 2025.

The AI segment, which builds AI models and runs the X platform, sells advertising on X, premium X and Grok subscriptions, data licensing, API access to Grok models, and cloud compute. AI Solutions & Infrastructure went from $311m in the second quarter of 2025 to $2,194m in the second quarter of 2026.

Table 1.1 Revenue by segment and revenue line, audited years

Source: 424B4 0001628280-26-042639, Note 3 Revenue and Note 19 Segments, for the three dollar columns. All products revenue is attributable to the Connectivity segment. The FY2025 share column, the FY2025 growth column and the three segment operating margin rows are this report's calculations from those disclosed figures; the filing does not print them.

Consolidated revenue compounded at 34.1% a year across the two audited transitions, 34.9% then 33.2%. The composition of that growth changed completely. Consolidated revenue rose $3,628m in 2024: the Connectivity segment added $3,730m of that and the AI segment subtracted $341m. In 2025 consolidated revenue rose $4,659m, of which the Connectivity segment added $3,788m. The Space segment contributed $239m and then $290m, from a base that started larger than the Connectivity segment's and finished at a third of its size.

The 2026 quarters break the pattern again.

Table 1.2 Revenue by segment, quarters and half years, unaudited

Source: 424B4 0001628280-26-042639 for the 2025 first quarter; 10-Q 0001628280-26-052535 for every June column.

Second quarter revenue rose 91.9% year on year. Management attributes $1,824m of the $3,743m increase to new AI infrastructure contracts, $1,703m to Starlink consumer and enterprise growth, and $216m to the Space segment on customer mix and one additional customer launch. Space segment revenue for the half fell 1.9%. Advertising fell 18.4%.

Geography

Revenue by geography is disclosed, on the basis of the country of domicile in which the transaction originated, and only for the three audited years. No 2026 period carries a geographic split: it is absent from Note 3 and Note 18 of the interim statements inside the prospectus, and from Note 3 and Note 18 of the 10-Q. Substantially all long lived assets sat in the United States at both audited balance sheet dates.

Table 1.3 Revenue by geography, audited years

Source: 424B4 0001628280-26-042639, Note 19 Segments.

The international share moved from 28.1% to 30.6% over two years. The basis is the domicile in which the transaction originated, so the table reflects where transactions were contracted, and an Ireland line of $1,827m says nothing about where those subscribers sit. The operating disclosure is wider than the accounting one: Starlink served 164 countries, territories and other markets at 31 March 2026 and 167 at 30 June 2026, and Starlink Mobile reached about 7.4 million monthly unique devices across about 30 countries through some 30 mobile network operator partners.

Cadence, subscribers and compute

Table 1.4 Key operating metrics

Source: 424B4 0001628280-26-042639 and 10-Q 0001628280-26-052535, Key Business Metrics. Starship counts for the annual columns are the residual between the total launches metric and the Falcon footnote; the 10-Q prints Starship separately for the June periods. Half year subscriber columns are the count at 30 June. Launch Services revenue per customer launch is not a disclosed metric: the filings give the two inputs and this report divides one by the other.

Total launches rose 73% across the two audited years while customer launches went from 33 to 45 and then fell back to 43, so the customer share of Falcon activity fell from 34.4% to 26.1% and to 22.1% in the first half of 2026. Internal Starlink deployment absorbed the growth: 86% of the 2,213 metric tons SpaceX put in orbit in 2025 was its own payload. Launch Services revenue per customer launch has been stable near $60m throughout, so the price per launch is holding and the number of launches is what varies.

Starlink is growing on volume while its price falls. Subscribers doubled year on year in each of the last four measured periods, reaching 12.0 million at 30 June 2026 across a constellation of over 10,200 broadband and mobile satellites (10-Q 0001628280-26-052535). The share of all satellites in orbit is a separate and earlier measurement: the prospectus puts the constellation at approximately 9,600 broadband and mobile satellites at 31 March 2026, approximately 75% of all active manoeuvrable satellites in orbit at that date (424B4 0001628280-26-042639). No later filing restates that share. Monthly ARPU fell from $99 in 2023 to $66 in the second quarter of 2026, a third lower, and management expects further decline as the base shifts outside North America and cheaper plans are added. Hardware moved the same way. Products revenue, all of it Starlink kits, was $1,470m in 2024 and $1,510m in 2025, against 2.1 million and then 4.5 million net subscriber additions. Kit revenue per net addition therefore halved, from about $700 to about $336.

Nameplate compute draw, the installed GPU power rating, reached 1.4 gigawatts at 30 June 2026 against 0.4 a year earlier, with COLOSSUS and COLOSSUS II online. That capacity is what the new AI infrastructure contracts sell.

Customer concentration

The filings quantify the government share once. Item 1A of the prospectus, under the risk factor on supplying services to the United States government, states that in 2025 approximately one fifth of revenue was attributable to agencies within the US federal government (424B4 0001628280-26-042639). On 2025 revenue of $18,674m that is roughly $3.7bn. Chapter 8 works the same disclosure against the concentration note.

No filing splits government from commercial revenue by segment, and none gives a government share for any 2026 period: neither figure appears in Note 3 of the prospectus, which carries both disaggregation tables and the concentration disclosure, in Note 19 Segments, in the MD&A under Components of Results of Operations or Segment Results, in the Business section under Government Contracts, or in Notes 3 or 18 of the 10-Q. Four further disclosures bound the question.

Table 1.5 Customer concentration, per cent of consolidated revenue

Source: 424B4 0001628280-26-042639, Note 3 Concentration of risk, and 10-Q 0001628280-26-052535, Note 3. Neither customer is named. Customer A revenue relates to all three segments; Customer B relates to the AI segment. The Launch & Development row is calculated from the disclosed revenue lines.

First, one unnamed customer supplied 20.9% of 2025 revenue and 17.9% of first half 2026 revenue, spanning all three segments; that is the same order of magnitude as the one fifth federal figure without being formally the same disclosure, because the note does not name the customer. Second, Launch & Development is government work by the filing's own definition, and it fell from 15.3% of consolidated revenue in 2023 to 4.8% in the first half of 2026, entirely because the denominator grew. Third, Enterprise & Government inside the Connectivity segment blends commercial aviation and maritime accounts with Starshield, so it cannot be split. Fourth, a second unnamed customer appeared in 2026 at 19.5% of second quarter revenue, in the AI segment, which is the new compute concentration.

The qualitative record is stronger than the quantitative one. SpaceX describes itself as the primary launch provider for the United States government. In 2025 it flew 11 of the 12 National Security Space Launch medium and heavy lift missions and all five United States crew and cargo missions to the International Space Station for NASA. It holds direct contracts with NASA, the Department of War, the General Services Administration and certain Intelligence Community agencies; the contracting terms and how they differ from the defence primes are in chapter 4. Named commercial and institutional customers include the National Reconnaissance Office, Axiom Space, SES, Eutelsat and OneWeb on the launch side, and United Airlines, Carnival, Maersk and John Deere on the Starlink enterprise side.

Contracted revenue points the same way. Backlog was flat through the first quarter of 2026 and then rose 71.8% in a single quarter, adding $19,840m, while the share expected within one year jumped from 36% to 56% (Table 3.7). The cloud services agreements behind that increase are described as monthly fee arrangements terminable by either party on 90 days notice after an initial ramp. The backlog is larger and shorter dated than it was three months earlier.

How the company was capitalised before listing

Twenty one Form D and Form D/A notices sit under CIK 1181412 between March 2009 and August 2022. Taking the latest notice against each SEC file number gives fifteen distinct offerings, $9,696m offered and $9,034m sold, on Rule 506 exemptions, with every notice declining to disclose issuer revenue.

Table 1.6 Private placements noticed on Form D

Source: Forms D and D/A filed under CIK 1181412, accessions 0001181412-09-000003 to 0001181412-22-000003. Rows headed by a D/A accession supersede the original notice against the same SEC file number. The November 2021 notice is the only one flagged as a business combination transaction.

Form D does not name investors. It is a notice of an exempt offering, and it has no field for the identity of a purchaser. Its Item 3 names related persons, meaning the issuer's own executive officers, directors and promoters; Items 12 to 16 give the size of the offering, the amount sold, the minimum investment accepted, whether any non accredited investors took part, how many investors have already invested, and any sales commissions and finders' fees. None of the twenty one notices supplies a buyer's name, and what they do disclose is identical across all of them: the minimum investment accepted is $0, no filing reports a non accredited investor, sales commissions are $0, finders' fees are $0 and the sales compensation item is left empty. Every related person named in thirteen years is a SpaceX insider. Elon Musk, Gwynne Shotwell and Luke Nosek appear on all twenty one notices, and Kimbal Musk on every notice through 2021. Steve Jurvetson and Antonio Gracias appear as directors from 2009 and 2010, and Donald Harrison from August 2017. Bret Johnsen appears on the 2015 notice and on every notice from April 2018, but on neither of the two 2017 notices, which name David Harris in his place. Tim Hughes, the chief counsel, is named in 2009, 2010 and 2015; the rest of the operating roster of engineering and legal officers is named only in the 2009 and 2010 notices. Who supplied the $9,034m is therefore not in this record at all. Chapter 6 sets out the four documents in which the filing record does name investors, and what each one gives.

The first two notices, in 2009 and 2010, sold $65m between them. The 2015 round sold $1,000m to thirteen investors. From 2020 the investor counts jump into the seventies and nineties as the constellation build and the Starship programme ran together, and $6,069m of the total $9,034m was sold in the three years from 2020 to 2022. Then the record stops: the last exempt offering was noticed in August 2022, almost four years before the listing.

The June 2026 offering raised $85,675m net, nine and a half times everything the Form D record shows the company raising privately in thirteen years. Chapter 5 sets out the mechanics. It arrived in a business whose consolidated operating result for 2025 was a loss of $2,589m and whose capital expenditure that year was $20,737m.

What the revenue model now depends on

Three dependencies fall out of the disclosure. Starlink consumer revenue is a subscriber count multiplied by a price management expects to keep falling, so growth requires net additions to keep doubling into markets that will pay less. Space segment revenue depends on a customer launch count that has stagnated near 40 a year while the fleet flies four times that number, and the segment moved from breakeven to a 16.1% operating loss in 2025 as Starship development was expensed. AI segment revenue is now the fastest growing and the most concentrated. A single customer is 19.5% of consolidated second quarter revenue, on contracts described as cancellable on short notice, and the compute capacity behind that revenue consumed $12,727m of capital expenditure in 2025 alone.

2 Financial Analysis & Ratios

SpaceX has never filed an annual report. There is no 10-K. The audited financial record consists of three income statement years and two balance sheet dates, and all of it sits inside one document, the 424B4 prospectus of 12 June 2026 (0001628280-26-042639). One 10-Q follows it, for the quarter ended 30 June 2026 (0001628280-26-052535). PricewaterhouseCoopers LLP, auditor since 2012, signed the three years to 31 December 2025. Everything after that date is unaudited.

A five year ratio series cannot be built and is not attempted here. Where a ratio needs an average balance across two consecutive year ends, only fiscal 2025 supports one. Where a ratio needs a pre IPO capital structure to mean anything, it is named in Table 2.9 with the reason for excluding it.

Revenue is growing quickly and the rate is rising: 34.9% in 2024, 33.2% in 2025, and 91.9% in the June 2026 quarter against the same quarter of 2025. Gross margin widened from 41.2% in 2023 to 55.3% in the June 2026 quarter. And operating income is negative all the same, because research and development spending rose 149.5% in 2025 and doubled again in the first half of 2026, overtaking cost of revenue as the largest item in the cost base. In three weeks in June the company raised $85,675 million of net IPO proceeds and $25,000 million of senior notes, converted $38,752 million of preferred stock into common equity, and moved from $1,851 million of net cash to $60,645 million.

What exists to analyse

Table 2.1 The filed record, in full

Three audited years and two audited balance sheet dates exist. The prospectus index and the audit opinion both state the coverage in those terms.

The statements

Table 2.2 Consolidated statement of operations, USD millions

Fiscal 2024 is the only profitable period in the record, and it is profitable partly on a $985 million other income line that includes a $955 million unrealised gain on digital assets. The 2023 operating loss carries a $3,775 million intangible impairment. Adjusted EBITDA is the company's own non GAAP measure, defined as net income excluding depreciation and amortisation, share based compensation, restructuring, impairments, interest, other income and expense, and tax; the prospectus and the 10-Q both set out how it is built back to net income. Chapter 3 works through what the excluded items are worth.

Table 2.3 Consolidated balance sheet, USD millions

Common equity was 2.8% of total assets at the end of 2025 and 66.0% six months later. The swing is a reclassification as much as a fundraising: $38,752 million of redeemable convertible preferred stock sat outside shareholders' equity as temporary equity until the IPO converted it. Property, plant and equipment grew 54.3% in six months, from $42,602 million to $65,736 million.

Table 2.4 Cash flow, USD millions

The last row is this report's derivation. Neither the prospectus nor the 10-Q presents a free cash flow number; the prospectus uses the phrase only to describe a stock option vesting condition. Operating cash flow has been positive in every filed period. Capital expenditure has exceeded it in every period since 2023, and by $25,010 million in the first half of 2026 alone.

Table 2.5 What changed

Ratios, and the ones the record will not carry

Table 2.6 Margins, every period the filings support

The quarter and the half year point in opposite directions, and one line explains the difference. Operating cash flow is not presented for the quarter alone in the 10-Q, so those two cells read n/a. The June quarter came within $143 million of operating breakeven on a 45.3% adjusted EBITDA margin. The half is dragged down by $1,962 million of other expense, of which $1,545 million is loss on extinguishment of debt: $1,526 million in March, when the bridge loan retired the inherited X and xAI facilities, and $18 million in June, when the notes retired the bridge loan.

Table 2.7 Growth, intensity and coverage

Table 2.8 Balance sheet ratios, at the three dates that exist

Return on average assets exists for one year only, because averaging needs two consecutive balance sheet dates and the prospectus supplies two dates in total. Working capital days for fiscal 2025 come to 26 days of receivables and 85 days of inventory against 313 days of payables, and the payables figure does not measure payables. $7,088 million of capital expenditure sat in accounts payable at 31 December 2025, 60% of the balance, so the numerator and the denominator are measuring different things. The cash conversion cycle is therefore excluded and explained in Table 2.9 instead.

Table 2.9 Ratios not presented, and the reason

Capital structure after the IPO and the notes

The two June financings are set out in chapter 5; this chapter takes what they did to the balance sheet. Six separate secured and inherited loan facilities became one set of unsecured notes in six months: the $20,000 million bridge loan signed on 2 March 2026 repaid the X and xAI loans, and the $25,000 million of senior notes issued on 26 June repaid the bridge loan.

Table 2.10 The debt stack, before and after

Interest on the notes is payable each 15 January and 15 July, first payment 15 January 2027, and the tranche terms are in Table 5.3. The effective interest rate at 30 June 2026 was 6.030%. The notes carried $24,852 million net of $148 million of deferred financing costs and were marked at a $24,697 million fair value at the same date. Annual cash coupon on the weighted average rate is about $1,464 million, against $1,293 million of total interest expense in the first half of 2026.

The revolving credit facility was increased from $1,500 million to $5,000 million in May 2026 and was undrawn at 30 June 2026, with all $5,000 million available. The financial maintenance covenant is a consolidated leverage ratio of no more than 3.75 times, measured on funded debt net of 85% of unrestricted cash, so the $93,522 million cash balance leaves the company far below that limit. Gross debt of $39,364 million against derived trailing twelve month adjusted EBITDA of $8,305 million is 4.7 times, against 3.5 times on the audited 2025 figures. Measured net of cash the company holds $60,645 million more cash than debt, and the ratio has no meaning.

Cash

Cash and cash equivalents went from $24,747 million at 31 December 2025 to $93,522 million at 30 June 2026, with a further $6,487 million in short term marketable securities. On the cash and restricted cash basis used in the statement of cash flows the move is $25,124 million to $94,352 million. The IPO contributed $85,675 million and debt raised net of repayments, issuance costs and extinguishment costs contributed $10,966 million; capital expenditure and intangibles took $29,352 million back out. Operating cash flow of $3,466 million was a tenth of what the two financings brought in. The 10-Q names cash, marketable securities, operating cash flow and the credit facility as the primary liquidity sources, and states that no material debt principal falls due until 15 July 2031.

At 28 July 2026 there were 7,696,293,669 Class A and 5,485,486,276 Class B shares outstanding, 13,181,779,945 in total, against a weighted average of 5,864 million shares in the June quarter and 2,926 million across fiscal 2025. Per share comparisons that cross the IPO are comparisons across two different share counts, and the loss per share of $0.09 for the June quarter against $0.34 a year earlier reflects that as much as it reflects the narrower loss.

3 MD&A & Management Commentary

The management commentary record at Space Exploration Technologies Corp. is ten documents long and one quarter deep. There is no annual report, no earnings call transcript on EDGAR, and no number that management has told the market to expect. Three of the ten documents carry the substance: a 424B4 prospectus that quantifies two quarters and three audited years driver by driver, a single 10-Q that repeats the same calculations for the June quarter and drops almost every forward looking sentence the prospectus contained, and an Exhibit 99.1 earnings release that adds the two contract awards which explain why backlog rose 72 per cent in ninety days.

Management has committed to engineering and corporate milestones and has published no financial target. Nothing management reported in the prospectus was restated by the 10-Q or the earnings release: the June quarter numbers join the March quarter numbers without an adjustment anywhere in the statement of operations, the segment tables, the workings behind Adjusted EBITDA, the capital expenditure table, the cash flow statement or the operating metrics. The single quantified expectation the prospectus set for profitability was a multiyear wait for sustained positive AI Segment Adjusted EBITDA. Seven weeks later that measure came in at a positive $1,146 million, which settles the sign and leaves the word sustained open. That is the commentary record in full.

3.1 The commentary record, and what each document is for

Table 3.1 is the whole population: three documents that carry a management discussion and seven current reports that carry management's own account of a single event. There is no eleventh document.

Table 3.1 The management commentary record as at 5 September 2026

Three sentences in that record are worth reproducing exactly, because no paraphrase preserves what they carry. The first is the company's own account of how it will speak to the market, Item 7.01 of the 8-K of 15 June 2026:

The Company does not intend to release quarterly or annual financial results or other material news through wire distribution services, such as Business Wire, PR Newswire, or similar services.

That is accession 0001628280-26-043288, Item 7.01. The channel is the investor relations page and the X account, and an investor waiting for a newswire will be told late or not at all. The second is the only cash figure filed between the IPO and the 10-Q, furnished under Item 7.01 of the 8-K of 22 June 2026 to support the note offering:

On June 22, 2026, in connection with the proposed offering of the Notes (as defined below), the Company provided updated disclosure to prospective investors indicating that, as of June 19, 2026, it held approximately $100.8 billion in cash and cash equivalents.

That is accession 0001628280-26-044489, Item 7.01. It stands four days after the offering closed and eleven days before the quarter end, the filing warns that the 30 June figure may differ materially, and it did: the 10-Q reports $93,522 million. The third is the whole shape of the June quarter in one sentence, from Item 2 of the 10-Q, accession 0001628280-26-052535:

This increase was due to an increase in revenue from our AI segment of $1,824 million primarily from new AI infrastructure contracts, an increase in revenue from our Connectivity segment of $1,703 million as our Starlink consumer subscribers and enterprise and government customer base continued to grow, as well as an increase in revenue from our Space segment of $216 million primarily driven by customer mix and an additional customer launch.

One caution attaches to the 4 August exhibit. It is internally inconsistent about compute. The AI slide says capacity expanded to 1.4 GW, “up from 1.0 GW in Q1 2026 and 0.4 GW in Q2 2026”, while the table on the same slide puts 0.4 GW in the Q2 2025 column. The comparative period in the sentence is misstated; the three figures themselves agree with the 10-Q and with Table 3.6 of this report. Quotations in this report from that exhibit are taken from the accessibility text layer behind the slide images, which carries the filing's line breaks as double spaces; runs of whitespace have been collapsed to single spaces and nothing else has been altered.

The prospectus MD&A carries eight named sections that the 10-Q MD&A does not reproduce at all: Our Repeatable Business Model, Segments in Our Vertically-Integrated Innovation Engine, Our Capital Allocation and Funding Strategy, the segment income commentary, the Segment Adjusted EBITDA commentary, Capital Expenditures, Drivers of Our Performance, and Quantitative and Qualitative Disclosures About Market Risk as an MD&A subsection. Market risk moved to Item 3 of the 10-Q and shrank to five sentences. The rest simply went.

An investor reading only the 10-Q therefore gets the numbers without the framework, and every quantified ambition this company has published sits in a document dated 12 June 2026 that management is under no obligation to update. The 424B4 says as much: the forward looking statements in the prospectus speak only as of the date of the prospectus, and the company undertakes no obligation to update them unless required by law.

3.2 Results of operations as management explains them

Group revenue in the June quarter was $7,814 million, up $3,743 million or 91.9 per cent. Loss from operations narrowed to $(143) million from $(970) million. On the half year the direction reverses: revenue up $4,370 million or 53.7 per cent, but loss from operations wider by $1,143 million at $(2,086) million, and net loss wider by $3,281 million at $(4,817) million. Both statements are true and both come from the same table. The June quarter shows the operating result on its own. The half year also carries the March quarter's $1,876 million of other expense, almost all of it debt extinguishment costs and an unrealised loss on Bitcoin.

Table 3.2 Consolidated results of operations, US$ million

The March 2026 column is the prospectus column, read from the 424B4 of 12 June. The June columns are from the 10-Q and the earnings release. They join without an adjustment.

Management's attribution for the 91.9 per cent quarter accounts for the whole increase. The 10-Q assigns $1,824 million of the increase to the AI segment, $1,703 million to the Connectivity segment and $216 million to the Space segment, and those three account for the whole $3,743 million. Within each segment the named components also add up to the segment total. Figure 3.1 sets the increase out driver by driver, with nothing left over except a $25 million residual inside AI solutions.

Interest income at $340 million against $98 million shows the effect of the new cash. The 10-Q attributes the increase to interest earned on cash equivalents and marketable securities following the deployment of proceeds from the IPO and the SpaceX Notes issuance. Those proceeds landed on 15 and 26 June, so $340 million was earned with the new money in place for roughly two weeks of the quarter. Interest expense rose to $629 million from $411 million on the debt raised by the company and, in the March quarter, on the financing arrangements entered into by the AI segment. Of the $629 million, $327 million was related party interest, a disclosure the income statement carries on its face.

The tax line falls on a valuation allowance. Management attributes the $115 million reduction in the quarterly provision and the $123 million reduction on the half to a change in the valuation allowance on United States research and development credits during 2025. The prospectus sets out what happened: a partial valuation allowance was released in 2024 against $659 million of forecast credit utilisation, and in 2025 that benefit was reversed on cumulative pretax losses adjusted for permanent differences and other negative evidence.

3.3 Segment by segment, on management's own drivers

Table 3.3 Segment results, US$ million

The Space segment, which launches rockets and builds spacecraft. Revenue rose 29.0 per cent in the quarter and fell 1.9 per cent on the half. Management gives one reason for both: customer launches went from 9 to 10 in the quarter and from 21 to 17 on the half, with a favourable customer mix shift in the quarter on top. The dollar split is Launch Services up $158 million and Launch and Development up $58 million in the quarter, against Launch Services down $78 million and Launch and Development up $48 million on the half, the latter attributed to the timing of work performed on government contracts.

The segment's operating loss widened to $(542) million from $(369) million, and to $(1,204) million from $(439) million on the half, on Starship. Research and development rose $383 million in the quarter, of which management assigns $311 million to production and engineering costs and $73 million to launch and test costs. On the half the same two lines carry $653 million and $134 million of a $787 million increase. Cost of revenue was flat in the quarter, with higher customer and launch overhead of $42 million offset by a $43 million decrease in production related costs.

The segment's operating result is therefore being consumed by a development programme, and management says so in the earnings release: total Space segment costs and expenses were up $389 million year over year as research and development on Starship accelerated, a programme management believes will reduce the cost to orbit by 99 per cent or more relative to the historical average.

The Connectivity segment, which sells Starlink satellite broadband. It produces all of the group's segment operating profit and is the only segment with positive Segment Adjusted EBITDA in every period shown. Revenue rose $1,703 million or 65.8 per cent, split by management into $764 million from consumer subscribers and $939 million from government, aviation, maritime and other enterprise business. The consumer half splits further into 101.2 per cent growth in subscribers offset by a 22.4 per cent decline in ARPU. On the half the same split gives $1,420 million consumer and $1,066 million enterprise and government, with 101.2 per cent subscriber growth against a 22.6 per cent ARPU decline. The earnings release adds the split in dollars: consumer revenue up 44 per cent year over year to $2,485 million, enterprise and government up 108 per cent to $1,806 million.

Enterprise revenue is now growing more than twice as fast as consumer revenue and reached 42 per cent of segment revenue in the June quarter, against 34 per cent a year earlier. Management names the causes in the earnings release rather than the 10-Q: an agreement with American Airlines, service activation with Southwest, Virgin Atlantic, Iberia and Aer Lingus, mobile partnerships with SoftBank, NTT Docomo and Spark NZ, and over $6 billion of multiyear United States government contracts for Starshield, primarily from two Space Force contracts for low Earth orbit communications and sensing constellations.

Cost of revenue rose $659 million on three named items: depreciation up $226 million from capitalised launch and satellite costs, operating expenses up $158 million, and Starlink Kit production spend up $148 million. Inside the operating expense figure management itemises customer support and installation of $50 million, ground operating costs of $44 million and payment processor fees of $17 million. Selling costs rose $160 million, of which marketing is $111 million, international expansion $16 million and sales and property taxes $15 million, partly offset by $13 million lower bad debt expense. Research and development rose $151 million, split $90 million next generation satellites, $28 million ground equipment and $18 million Starlink Kits.

The AI segment, which builds AI models and runs the X platform. Revenue rose 247.5 per cent to $2,561 million. Management attributes $1,883 million of the $1,824 million net increase to AI solutions and infrastructure, offset by a $59 million fall in advertising, and within the first figure names $1,600 million of AI infrastructure revenue from the start of cloud services and $258 million from Grok and X subscriptions. The earnings release supplies the contract source: several Cloud Services Agreements totalling $14.1 billion of contracted sales, of which $1.6 billion was recognised in the quarter.

Management attributes the advertising line to a platform change. The 10-Q states that the decrease in advertising revenue was due to the Company's transition to a new advertising platform which impacted ad sales for a short period of time. Advertising revenue was $367 million in the June quarter against $426 million a year earlier and $343 million in March, so the quarter sequentially recovered $24 million while remaining $59 million below the prior year. On the half, advertising is down $160 million. Management also flags advertising seasonality in both filings, with spend highest in the fourth quarter and lowest in the first.

Costs rose $1,557 million, dominated by research and development up $1,056 million on infrastructure and cloud computing of $726 million and employee compensation of $180 million. Cost of revenue rose $555 million, of which $470 million is infrastructure and cloud computing reallocated to cost of revenue as infrastructure revenue arrived. Selling costs rose $134 million, which management attributes primarily to $177 million of employee compensation, partly offset by a $64 million fall in legal expenses following a dismissal of litigation against the company. Those two items are named as the main causes rather than as the whole of the movement, and they do not close it.

Table 3.4 Revenue drivers management named and sized, US$ million

Table 3.5 Cost drivers management named and sized, US$ million

Customer support and installation, ground operating costs and payment processor fees are components of the Connectivity segment operating expense line above them. The same applies to the Grok and X subscriptions and AI infrastructure lines inside AI solutions in Table 3.4.

Read down the right hand columns of Table 3.5 and one item dwarfs the rest. Infrastructure and cloud computing carries $1,742 million of the $2,527 million rise in AI segment research and development on the half, plus a further $386 million in cost of revenue. Every other named driver in the whole company is an order of magnitude smaller. Figure 3.2 puts the quarter on one scale.

3.4 The line management calls Adjusted EBITDA, and the quarter it turned

Adjusted EBITDA and Segment Adjusted EBITDA are the measures management says it uses internally for planning and forecasting. Adjusted EBITDA excludes depreciation and amortisation, share based compensation, restructuring charges, impairments, interest expense, interest income, other income and expense, and income taxes. Segment Adjusted EBITDA starts at segment income or loss from operations and excludes the first four of those.

Depreciation and amortisation was $2,848 million in the June quarter against $1,526 million a year earlier, and $5,290 million on the half against $2,970 million. Share based compensation was $831 million against $463 million, and $1,470 million against $694 million. Together they add $3,679 million back to a $(541) million quarterly net loss. The AI segment alone contributes $1,885 million of the quarter's depreciation, which follows from $23,551 million of half year capital expenditure. An investor who treats AI Segment Adjusted EBITDA of $1,146 million as cash generation is reading past the segment's $(1,257) million operating loss and the depreciation schedule that produced the gap.

The prospectus was explicit on this measure and on this segment. Management wrote that it expects to continue expanding its terrestrial data centres, and to launch orbital data centres, and that it expects a multiyear investment horizon before these deployments translate into sustained positive Segment Adjusted EBITDA for the AI segment. The prospectus was filed on 12 June 2026. Seven weeks later the earnings release reported that the AI segment had cut its operating loss by 49 per cent against the March quarter to $1.3 billion and had achieved positive AI Adjusted EBITDA of $1.1 billion.

Two readings fit the same facts. One is that the Cloud Services Agreements arrived faster than the prospectus assumed, so the multiyear horizon closed early. The other is that the word management used was sustained, and one quarter carrying $1,600 million of new infrastructure revenue against $23,551 million of half year capital expenditure does not establish that. AI Segment Adjusted EBITDA for the March quarter was $(609) million. Management has not repeated the multiyear framing in any filing since 12 June, and has not withdrawn it either.

3.5 Key business metrics, and the hedge management added to one of them

Table 3.6 Key business metrics as management reports them

Mass to orbit fell 26 per cent in the quarter and 6 per cent on the half. Launches fell from 46 to 38 and from 84 to 78. Both declines sit inside a company whose revenue almost doubled, and the reason is the sentence management uses to introduce the metrics. In the prospectus it reads that mass to orbit and launches grow more rapidly than Space segment revenue, because the metrics include internal constellation deployments from which no intersegment revenue is recognised. In the 10-Q the identical sentence gained one word: mass to orbit and launches generally grow more rapidly than Space segment revenue.

That single adverb was added in the first quarterly report in which the claim was untrue in both directions: mass to orbit fell while Space segment revenue rose 29 per cent in the quarter, and mass to orbit fell 6 per cent while Space segment revenue fell 1.9 per cent on the half. The internal payload line is where the decline sits, down from 563 to 397 metric tons in the quarter, which is consistent with a constellation transitioning from Falcon deployment of V2 Mini satellites to Starship deployment of V3 satellites. Customer payloads were flat at 87 and 88 metric tons.

The prospectus states the direction for Falcon plainly. Management wrote that while it has steadily increased its Falcon 9 launch cadence over recent years, it expects Falcon 9 launches to decrease over time. Against 165 Falcon launches in 2025 and 134 in 2024, the 77 recorded in the first half of 2026 puts the year on a lower path, and 37 in the June quarter against 45 a year earlier confirms the direction inside three months of the statement being made.

ARPU is the second stated expectation, and the filed figures have followed it. The prospectus states that management generally expects Starlink Subscriber ARPU to continue to decline over the next few years as the base outside North America grows, as lower priced plans are added and as monthly fees are adjusted. It notes the fall from $91 per month in 2024 to $81 in 2025, and from $86 in the March 2025 quarter to $66 in the March 2026 quarter. The June quarter held at $66, which the earnings release records as maintaining ARPU at $66, in line with the first quarter of 2026. Subscribers doubled to 12.0 million over the year and added 1.7 million sequentially. The two figures management publishes for consumer revenue, 101.2 per cent more subscribers against 22.4 per cent less revenue per subscriber, are the whole Connectivity consumer model.

Nameplate compute draw is defined as installed GPU count multiplied by all in power draw, and management is careful to say it reflects installed capacity rather than consumption or utilisation, and excludes cooling, distribution losses, lighting, security and facility overhead. It went from 0.4 gigawatts to 1.4 gigawatts in a year, and the earnings release attributes the step to the continued buildout of Colossus II with significant incremental capacity under construction. The sentence that makes that attribution labels the 0.4 gigawatts as the second quarter of 2026 rather than of 2025. Its own table on the same slide assigns the figure to 2025. No utilisation figure accompanies any of them in any of the ten documents.

3.6 Backlog, the measure that carries across all three filings

Backlog appears nowhere in Item 2 of the 10-Q. The figures below come from the revenue note to the financial statements in the prospectus and in the 10-Q, and the only place management itself uses a backlog number in commentary is the CFO commentary of the 4 August earnings release, accession 0001628280-26-052515, where the June figure is given as $47.5 billion. It is the only management measure disclosed on three dates inside nine months.

Table 3.7 Backlog and its expected timing of recognition

The prospectus describes the December 2025 middle bucket as expected to be recognised in 2027 and 2028 rather than as a one to three year band, so that row is a period label rather than a duration.

Backlog fell $756 million in the March quarter and rose $19,840 million in the June quarter. Management names two awards that account for almost the whole move: $14.1 billion of contracted sales under Cloud Services Agreements and over $6 billion of Starshield contracts. Those two total roughly $20.1 billion against a $19.84 billion increase, and the difference is revenue recognised out of backlog in the quarter.

The composition change matters more than the level. Within one year backlog went from 32 per cent to 56 per cent of the total in six months. In dollars that is roughly $9.1 billion at December against roughly $26.6 billion at June, close to a threefold rise. The company is converting a slow, government weighted order book into a fast, compute weighted one. Management defines contracted sales in the earnings release as the total value of contracts signed for the period in which they cannot be cancelled and are enforceable, and states that the figure excludes estimated revenue for future periods that either party can cancel.

3.7 Liquidity and capital resources

Management's liquidity paragraph changed more between June and August than any other part of the MD&A, because the balance sheet changed more. At 31 March 2026 the company held $15,852 million of cash and $7,823 million of short term marketable securities, with $1,500 million available under an undrawn revolver and $29,132 million of debt principal outstanding, of which $20,000 million was a bridge loan maturing in September 2027. At 30 June 2026 it held $93,522 million of cash and $6,487 million of marketable securities, with $5,000 million available under the amended revolver and $38,433 million of debt principal, of which $25,000 million matures between 2031 and 2056.

Table 3.8 Liquidity as management states it

The listing. Chapter 5 sets out the offering mechanics. Two points belong to the MD&A. Underwriting commissions and offering costs of $575 million are 0.67 per cent of the $86,250 million of gross proceeds. And the prospectus had set out approximately $74.4 billion of net proceeds on the base offering of 555.6 million shares, and $85.7 billion if the underwriters took their option in full. The base figure implies about the same $575 million of cost, so the 83.3 million over allotment shares carried no additional issue expense. The proceeds landed at the top of the range the prospectus published.

The prospectus states the intended use of proceeds as funding the growth strategy, including expansion of AI compute infrastructure, enhancements to launch infrastructure and launch vehicles, increases in the scale and capacity of the satellite constellations, and any remaining amounts for general corporate purposes. It also warns that management will have significant flexibility in applying the net proceeds and cannot predict with certainty all the particular uses. Half year capital expenditure of $28,476 million splits into $23,551 million for the AI segment, $2,699 million for the Connectivity segment and $2,226 million for the Space segment, which follows the stated order of priority and its stated emphasis. What the prospectus does not say, and no later filing says either, is how much of the $85,675 million is earmarked to each. The 10-Q adds only that if near term data centre needs decrease in scale or ramp more slowly than expected, capital expenditure may be reduced in that segment and reallocated to others.

The notes. The three notes 8-Ks are set out in chapter 5, and the tranche terms are in Table 5.3. Two things in them are MD&A material. The registration rights agreement commits the company to use commercially reasonable efforts to complete a registered exchange offer no later than 540 days after the issue date, which puts that deadline in December 2027. And the 10-Q describes the applicable spreads as between 0.002 and 0.003 basis points, where the 8-K of 26 June sets them out as plus 20 to plus 30 basis points; the 8-K figures are the ones consistent with a par call construct.

The refinancing traded a higher interest rate for a longer maturity. Management last disclosed an effective rate of 4.58 per cent on the bridge loan at 31 March 2026 and discloses 6.03 per cent on the notes at 30 June 2026, a step of about 145 basis points. In exchange the first material principal maturity moved from 28 August 2027 to 15 July 2031, and Item 3 of the 10-Q records that the company had no variable rate debt outstanding at 30 June 2026. At the weighted average coupon of 5.855 per cent the notes carry roughly $1,464 million of annual interest. Half year Adjusted EBITDA of $4,665 million covers half year interest expense of $1,293 million 3.6 times.

The prospectus stated a matching aim in the capital allocation section: management plans to access a range of debt and equity financing solutions available to it as a public company and aims to maintain an investment grade credit rating. The notes covenant package is consistent with that aim. The 10-Q describes restrictions on additional liens above 7.5 per cent of consolidated total assets and on fundamental changes, and the revolver carries a consolidated leverage ratio covenant of no greater than 3.75 to 1.0, stepping temporarily to 4.25 to 1.0 after certain qualified acquisitions. Management states it was in compliance with all covenants at both dates.

The cost of getting out of the old debt. The half year carries $1,545 million of loss on debt extinguishment. The 10-Q splits it: $1,526 million on the March repayment of the X B-1 Term Loan, the X B-3 Term Loan, the xAI Fixed Rate Loan, the xAI Floating Rate Loan and the xAI 12.5% Senior Secured Notes, and $18 million on the June repayment of the bridge loan. Note 9 gives the prepayment penalties instrument by instrument rather than as one number: $425 million on the two X term loans and $221 million on the two xAI term loans, both repaid on 2 March, and $518 million on the xAI 12.5% notes, repaid on 5 March. The prospectus states the aggregate in Note 21, at $1,163 million. Financing activities carry $1,153 million of debt extinguishment premium paid in cash. Adding a $539 million unrealised loss on Bitcoin, whose fair value fell from $1,637 million to $1,098 million against a $661 million cost basis on 18,712 units, the two are $2,084 million of charge against $1,962 million of net other expense on the half. The $122 million difference is foreign exchange and other gains.

Cash flow. Operating cash flow rose to $3,466 million from $351 million. Management attributes the $3,115 million improvement to lower net loss excluding noncash items and a $1,489 million working capital benefit from deferred revenue on upfront payments from Space segment and Connectivity segment customers, offset by a $1,533 million build in receivables and a $397 million decrease in payables. Deferred revenue rose to $14,286 million from $12,116 million over the half, and receivables to $3,596 million from $1,579 million.

Investing outflow of $34,487 million is $28,476 million of capital expenditure, $6,382 million of net purchases of marketable securities and $856 million paid to EchoStar for spectrum, against $1,195 million received from product rebates and $32 million of other items net. Financing inflow of $100,291 million is $85,675 million of IPO proceeds, $51,812 million of debt and other financing proceeds and $8,635 million of equity issues and awards, against $40,673 million of repayments, premium and issuance costs, $4,985 million of repurchases and settlement taxes and $173 million of finance lease principal.

How long the cash lasts. Management's statement is unchanged in meaning and shorter in form. The prospectus says the company believes it has sufficient sources of funding to meet its business requirements for at least the next twelve months from the issuance of the consolidated financial statements. The 10-Q says the same for at least the next twelve months, dropping the issuance date qualifier. Neither document says how long the money lasts. That can be bounded from what is filed: $100,009 million of cash and marketable securities plus $5,000 million undrawn under the credit facility, against half year capital expenditure of $28,476 million and half year operating inflow of $3,466 million. At the June quarter's rate of $18,369 million of capital expenditure a quarter, and with no contribution from operations, the balance sheet alone funds more than five quarters. At the half year average of $14,238 million a quarter it funds seven. Both are longer than the twelve months management asserts.

Table 3.9 Committed and contingent cash management identified

One item in Table 3.9 is fixed in dollars rather than in shares. The $42.40 share price for the spectrum consideration was set in an agreement signed on 7 September 2025 and amended on 5 November 2025, against an offering price of $135.00 in June 2026. The 261.8 million shares are issued at the Spectrum Acquisition Closing, expected 30 November 2027, and the fixed dollar value means the share count does not fall if the market price rises.

3.8 The guidance record

There is no financial guidance. No revenue, earnings, margin, Adjusted EBITDA or capital expenditure figure for any future period appears in the 424B4 MD&A or in its Cautionary Statement Regarding Forward-Looking Statements, in Item 2 of the 10-Q, in Item 2.02 of the 8-K of 4 August or its Exhibit 99.1, in the three senior notes 8-Ks of 22, 23 and 26 June, or in the 8-Ks of 15, 16, 17 June and 14 August. Those ten documents are the entire management commentary record and each was searched for it. The word guidance appears in the prospectus in accounting and tax contexts, in a FATCA passage about Treasury regulations and administrative guidance, and once in an engineering sense, in the risk factor listing separation events or guidance, navigation and control errors among the causes of launch failure. In none of them does it carry a financial forecast. The word outlook appears there only in the titles of cited third party market studies, in a risk factor about credit ratings, and in the list of words that signal a forward looking statement.

Management has committed instead to twenty nine engineering and corporate milestones, most of them dated and nineteen of them in one document. Table 3.10 is the full record with the filing and date attached to each, and the later filing that settles it where one exists.

Table 3.10 The guidance record

Quotations in Table 3.10 are reproduced exactly as filed, including each filing's own date forms, hyphenation and punctuation. Those taken from the earnings release come from the accessibility text layer behind the ten slide images, in which runs of whitespace have been collapsed to single spaces and nothing else altered.

Nineteen of the twenty nine statements were made on one day in one document. Twelve remain open, and the only two open items carrying a hard date are the orbital compute deployment in 2028 and the exchange offer deadline in December 2027. Seventeen are settled or partly settled by a later filing, and none of those went against the company inside the observation window. Item 10 is the one that is only partly settled: the positive AI Segment Adjusted EBITDA of the June quarter settles the sign of the number and leaves the word management actually used, sustained, still open. Item 7 moved both ways. The Launch and Development share of Space segment revenue rose on the half but fell on the quarter, from 34.3 per cent to 32.6 per cent, because the June quarter carried a Launch Services figure of $648 million.

That record of settled statements carries a caveat the prospectus itself supplies. Management wrote that forecasts, goals, milestones and expectations covering multiyear time horizons or unknown timelines inherently involve increased risks with respect to predictability, and that actual results may differ materially. It also warned that in rapidly evolving and highly competitive markets it may from time to time rapidly adjust, modify or change strategic priorities, capital allocation, product focus or operational initiatives. Twelve weeks of settled milestones on physical programmes with existing hardware is not evidence about a 100 gigawatt orbital compute goal.

3.9 What management changed between the prospectus and the 10-Q

Nothing management reported was restated. The three months ended 31 March 2026 and 31 March 2025 are the only periods common to the prospectus and the later filings. Both reproduce exactly from the 10-Q's six month and three month columns and from the earnings release's March 2026 column, across the full statement of operations, all three segment profit and loss statements, the workings behind Adjusted EBITDA and Segment Adjusted EBITDA, the capital expenditure table, the cash flow statement, the launch and mass to orbit metrics and the Space segment revenue mix percentages. A handful of March 2025 comparatives differ by $1 million and offset in pairs between the Connectivity and AI segments, which is rounding within a six month total rather than a change of view.

The language changed in three places. Table 3.11 sets the edits out with both versions verbatim.

Table 3.11 MD&A sentences that changed

Three sentences present in the prospectus liquidity section are absent from the 10-Q's. The first is worth having exactly as it was filed, because it is a claim about how this company is funded:

The cash we generate from our core operations also enables us to fund our research and development projects including our Starship rocket and next-generation satellites, the construction of future data centers, and the continued expansion of our AI-enabled products.

That is accession 0001628280-26-042639, MD&A, Liquidity and Capital Resources, and it has no counterpart anywhere in the 10-Q. The second said management continually evaluates its cash needs and may decide it is best to raise additional capital or seek alternative financing, including through drawdowns on existing or new facilities. The third said management may seek to refinance the SpaceX Bridge Loan, including with the proceeds of notes offerings, bank borrowings or other financial arrangements.

The third sentence no longer applies: the bridge loan was refinanced. The second is a qualification the company no longer needs with $100 billion on hand. Half year operating cash flow of $3,466 million against half year capital expenditure of $28,476 million means core operations funded 12 per cent of the investment programme. Removing the claim that operations fund the development programme is consistent with those two figures.

One further point of construction affects every comparative in this chapter. Beyond the common control combination set out in chapter 1, the prospectus states that the consolidated financial statements were retrospectively recast for all periods presented to reflect a five for one stock split effective 4 May 2026. The AI segment therefore has a 2023 operating loss of $(3,973) million in a company that did not own it until 2026.

3.10 Trends and risks management flags in the MD&A itself

These are separate from Item 1A. Each is a statement management chose to place inside the discussion of results or liquidity, where it bears directly on a number.

Table 3.12 Trends management identifies in its own discussion

The fourth row carries management's own account of how investment timing bears on results. Speed is treated as a competitive advantage, investment is accelerated when technical progress or market opportunity allows, and operating results are expected to fluctuate as a consequence. Management states that policy in the MD&A under Drivers of Our Performance rather than in the risk factors (424B4 0001628280-26-042639).

The eighth row is the one with a number attached. AI capital expenditure went from $749 million in the June 2025 quarter to $15,828 million in the June 2026 quarter, a factor of twenty one in a year, and management's stated mitigant is the ability to slow it and move it elsewhere. That flexibility is asserted in both the prospectus and the 10-Q and has never been tested in a filed period.

3.11 Critical accounting estimates where they bear on the numbers

The 10-Q refers back to the prospectus and states that there have been no material changes to the critical accounting policies and estimates as described in it. The prospectus is therefore the only source.

Table 3.13 Critical accounting estimates and the sensitivities management disclosed

The satellite useful life sensitivity is the one that moves the reported numbers most. A one year change in average satellite life is worth about $480 million of 2025 operating income against a 2025 operating loss of $(2,589) million, and about $170 million of March 2026 quarter operating income against a quarterly operating loss of $(1,943) million. Connectivity segment depreciation is running at $805 million a quarter and rose $503 million on the half from capitalised launch and satellite costs, so the estimate governs the segment that produces all of the group's operating profit. Management's stated inputs for it, particularly the evolution of constellation density and technology, are exactly the variables a Starship enabled V3 deployment changes.

The reusable vehicle estimate works differently. Boosters, fairings and Merlin engines are classified as property, plant and equipment and depreciated over the expected number of average flights rather than over time, against a maximum accounting useful life the prospectus sets at 25 flights. Second stages are not reusable and go straight to cost of revenue. That is why Space segment cost of revenue fell $189 million in 2025 on increased reusability, and why a Falcon fleet flying less often as Starship takes over feeds directly into the flights estimate.

Market risk carries three disclosed sensitivities and one of them has been extinguished. A 10 per cent adverse move across all currencies would have produced an immaterial gain or loss at 31 March 2026 and 31 December 2025, and the company does not hedge. A 100 basis point move in rates would have moved interest income immaterially for 2025 and the March 2026 quarter. A 100 basis point increase would have raised annual interest expense by approximately $200 million on the bridge loan at 31 March 2026. That last figure no longer applies: Item 3 of the 10-Q states that the company had no variable rate debt outstanding at 30 June 2026. Interest rate risk has moved from the company's borrowings, where the rate was floating and the amount material, to its cash: $100,009 million of cash and securities is now the exposure, and every basis point is worth about $10 million of annual interest income.

3.12 What management said about itself

The earnings release carries a section headed CFO Commentary, and the notes 8-K of 26 June is signed by Bret Johnsen as Chief Financial Officer. It is the only place in the record where management characterises the quarter in its own voice rather than explaining a movement.

Three sentences carry the message. The first frames the period, calling 2026 a momentous year so far and saying the second quarter demonstrated the true power of SpaceX. The second frames the balance sheet: the company ended the second quarter with $100 billion of cash, cash equivalents and marketable securities, and $47.5 billion in backlog. The third is the operating claim, that revenue growth accelerated across all business segments and the company delivered strong operating leverage, with significant margin expansion led by its new AI compute agreements.

The operating leverage claim is testable and it holds at the quarter. Revenue rose 91.9 per cent while total costs and expenses rose 57.8 per cent, and loss from operations narrowed from $(970) million to $(143) million. On the half it does not hold: revenue rose 53.7 per cent while costs rose 60.7 per cent and the operating loss widened. The claim is made about the quarter and is accurate about the quarter. The attribution of margin expansion to the AI compute agreements is also supported: $1,600 million of additional AI infrastructure revenue arrived against $470 million of additional AI segment cost of revenue.

None of the ten documents says what the next quarter or the next year will produce. The company has just raised $110,675 million and is spending $18,369 million a quarter on capital projects, and no capital expenditure plan for any future period appears anywhere in the record. Management describes what the spending is for, in what order of priority and how far it can be reversed, and gives its size only after the fact.

4 Sector & Competitor Analysis

There is no clean listed comparator for SpaceX, and no combination of listed companies makes one. The company reports three segments that belong to three different industries. The Connectivity segment, at 60.3 per cent of first half 2026 revenue, is a satellite broadband operator. The AI segment, at 27.0 per cent of first half 2026 revenue, builds AI models and runs a social platform. The Space segment, which launches rockets and builds spacecraft, is 12.6 per cent of first half 2026 revenue and fell 1.9 per cent year on year. A peer set built for any one of those segments misprices the other two.

This chapter works on the six months to 30 June 2026 throughout, because that is the period on which every comparator reports. Chapter 1 works on the audited year to 31 December 2025 and on the June 2026 quarter, so its segment shares differ from the ones here by period rather than by measurement: the Connectivity segment was 61.0 per cent of revenue in FY2025 and 54.9 per cent in the June 2026 quarter, against 60.3 per cent in the half.

EDGAR itself records the difficulty. SpaceX files under SIC 7370, Services-Computer Programming, Data Processing, so a screen on the registrant's own industry code returns software companies. Rocket Lab and Lockheed Martin file under 3760, Guided Missiles and Space Vehicles; Iridium, Viasat, EchoStar and AST SpaceMobile under 4899, Communications Services. No code contains SpaceX and its competitors together.

This chapter therefore compares segment by segment.

Table 4.1 SpaceX by segment, six months to 30 June 2026

Source: Form 10-Q for the quarter ended 30 June 2026, segment note, accession 0001628280-26-052535, retrieved through the SEC-API.io MCP server. Revenue and the segment operating income behind the margin column are disclosed there; the share of revenue, the growth on H1 2025, the operating margin and the contribution to group growth are this report's calculations from those disclosures.

The Connectivity segment earned $2,844m of segment operating income in the half while the Space segment lost $1,204m and the AI segment lost $3,726m. Group revenue growth of 53.7 per cent is faster than the 33.2 per cent and 34.9 per cent the audited years recorded, and the acceleration comes from the Connectivity and AI segments. Space segment revenue is flat to lower.

The audited history behind that is short: three years, inside the prospectus and nowhere else, all of them retrospectively recast for the common control combinations described in chapter 1, so no filed period shows the launch and satellite business on its own.

Selecting the comparators

A company earns a place here if the prospectus itself names it as a competitor, if it is listed with SEC periodic reporting, and if it maps onto a segment carrying material SpaceX revenue. The prospectus names its own competitors by segment in Business, Competition, so the set follows the filing's own list.

Table 4.2 The comparator set and the dimension each one carries

Table 4.3 Candidates rejected, and why

The like for like comparison

Two things limit what can go in one table. Fiscal years differ: Viasat's ends 31 March, so its most recent quarterly report covers three months to 30 June 2026 rather than six, and Lockheed Martin closes on the last Sunday of the month. Each row states its own period. And gross margin is available for three names only. SpaceX, Rocket Lab and Lockheed Martin disclose a full cost of revenue. The other five all carry a cost line on the face of the income statement, but each is drawn on a basis that does not subtract the same things, so a margin computed from it would not measure the same quantity. EchoStar presents cost of services and, separately, cost of sales for equipment and other, both captioned exclusive of depreciation and amortisation, which for a satellite operator omits the largest cost it carries (0001104659-26-089370). Iridium presents cost of services on the same exclusive-of-depreciation-and-amortisation basis, alongside a separate cost of subscriber equipment (0001418819-26-000045). AST SpaceMobile presents cost of revenues for products and for services, both captioned exclusive of items shown separately below (0001193125-26-342550). Northrop Grumman presents operating costs and expenses split between product and service, with general and administrative expenses outside them on their own line (0001133421-26-000034). Viasat presents cost of service revenues and cost of product revenues but carries amortisation of acquired intangible assets on a separate operating line outside both (0001193125-26-337903). Those five read n/a rather than carry a number that would mislead.

Table 4.4 Operating comparison, most recent reported period

Source: each registrant's Form 10-Q, accessions in model.py, retrieved through the SEC-API.io MCP server. Only the revenue column is a disclosed figure. Growth, gross margin, operating margin, capital expenditure over revenue and net debt over equity are this report's calculations from the filed line items of each registrant. Net debt is borrowings less cash and short term investments, taking each registrant's borrowings as its own balance sheet presents them: for SpaceX that includes finance lease liabilities, and for Rocket Lab it includes both long term borrowings and convertible senior notes. AST SpaceMobile's growth is not meaningful: revenue rose from $1.9m to $46.3m, and it presents no operating income line. Net income is not compared because EchoStar's half year carries a $9,729m deconsolidation gain against $906m of operating income, so a net margin column would rank it first on an item that is not trading.

Capital intensity is where SpaceX separates from everything else. It spent $28,476m of capital in a half year in which it recognised $12,508m of revenue, $2.28 of capital per dollar of sales. Lockheed Martin and Northrop Grumman together spent $1,298m in the same period, so SpaceX outspent both primes combined by twenty two times while earning roughly a fifth of their combined revenue. Only AST SpaceMobile spends more relative to revenue, and AST SpaceMobile has almost no revenue to spend against.

The balance sheet carries it. SpaceX holds $93,522m of cash and $6,487m of securities against $39,364m of debt and finance leases, net cash of $60,645m after $85,675m of net IPO proceeds and the June senior notes issue. Iridium sits at the other extreme, at 3.34 times net debt to equity, which is what a constellation looks like once it is built and financed with debt. Viasat and EchoStar sit between. SpaceX is funding its build from equity where the incumbents funded theirs from debt.

Segment against segment

Starlink is the most profitable satellite operator in the set. The Connectivity segment's 37.7 per cent operating margin in the first half of 2026 is roughly double Iridium's 19.1 per cent, three times EchoStar's 12.5 per cent and nine times Viasat's 4.1 per cent, on annualised revenue seventeen times Iridium's. That is the highest operating margin in the group.

The Space segment is the smallest of the three and the only one whose revenue fell. It lost $1,204m on $1,581m of revenue, while Lockheed Martin's Space segment alone booked $6,924m of sales at a 9.4 per cent margin in the same half, more than four times SpaceX's Space segment revenue. Launch cadence and launch revenue are different things: for a defence prime contractor, space means satellites, missile defence and strategic systems, and those are larger markets than launch.

The AI segment has no listed comparator at all. Its nearest named rivals are private, and the listed ones are not comparable in scale.

Valuation

Table 4.5 Valuation on the report price block, 3 September 2026 close

Prices and share counts are the report's single price block, struck at that close and reproduced in chapter 9. Prices are last sale prices from public market data. Share counts and their sources come from SEC filings retrieved through the SEC-API.io MCP server, and are from each registrant's own 10-Q cover page. Every column except the price is this report's calculation: market value is the closing price multiplied by the cover page share count, net debt is borrowings less cash and short term investments, enterprise value is the sum of the two, and annualised revenue doubles a half year and quadruples a quarter, so the last column is struck on a run rate and not on a trailing twelve month figure. None of the four is a disclosed figure. Three notes travel with the block. EchoStar trades under ECHO, not SATS: the cover of its 8-K filed 18 June 2026 still gives the trading symbol as SATS (0001415404-26-000027) and the cover of its 8-K filed 25 June 2026 gives ECHO (0001415404-26-000030), so the symbol changed between those two filings, and its 10-Q cover registers ECHO (0001104659-26-089370). No filing states the date of the change. A figure sourced under SATS after that week is stale. Viasat's 3 September close is $71.72, as read from market data. Viasat's fiscal year ends 31 March, so its quarter to 30 June 2026 is its first quarter of fiscal 2027.

SpaceX is worth six and a half times the seven comparators combined. At 76.5 times annualised revenue it is priced between Rocket Lab at 41.3 times and AST SpaceMobile at 268.5 times, and an order of magnitude above every profitable operator in the set. The accounts behind that multiple show research and development at 56.5 per cent of revenue in the first half of 2026 and operating cash flow less capital expenditure of $(13,952)m in 2025.

EchoStar's multiple carries a caveat. SpaceX is buying its AWS-3, AWS-4 and H block spectrum for $19.6bn, of which $11.1bn is SpaceX equity struck at $42.40 per share, so part of EchoStar's own equity value is a claim on SpaceX stock. Its enterprise value is not a clean read on a satellite operator.

The sector as the prospectus frames it

The prospectus puts its quantifiable addressable market at $28.5 trillion: $370bn for the Space segment, $1.6tn for the Connectivity segment and $26.5tn for the AI segment, excluding China and Russia. The components do not add up to that total. The two named Connectivity components, $870bn of Starlink Broadband and $740bn of Starlink Mobile, already exceed by themselves the $1.6tn segment total they sit inside, and the prospectus says the $1.6tn covers additional opportunities in enterprise and government as well, sizing business broadband at $200bn and government satellite communications at $5bn. Add the filing's own four components and the segment comes to about $1.8tn against the $1.6tn it prints.

How each figure was built matters more than its size. The $660bn consumer broadband figure is 1.8 billion global households multiplied by a weighted average $31 monthly ARPU, itself blended from $43 in high income markets, $16 in upper middle income and $9 in lower middle and low income. It assumes Starlink can address every household on Earth. The $370bn Space segment figure is a third party estimate of space enabled solutions and is the only one anchored to a market that exists at current prices. The $26.5tn AI segment figure is 93 per cent of the total, and $22.7tn of it is enterprise applications, a single line worth 80 per cent of the whole estimate. That category measures the software and labour AI might displace. Strip it out and the quantified opportunity falls to $5.8tn.

Government procurement

SpaceX contracts directly with NASA, the Department of War, the General Services Administration and Intelligence Community agencies. Two structural differences separate it from the defence primes. It is almost always the prime contractor and rarely uses subcontractors, and all of its launch contracts with US government agencies are firm fixed price with milestone based payments. Lockheed Martin, by contrast, recorded 41 per cent of first half 2026 sales on cost reimbursable contracts, and 70 per cent of its sales came from the US government. Under firm fixed price, cost reductions accrue to SpaceX; under cost reimbursable, they accrue to the customer. SpaceX therefore carries the cost overrun risk that cost reimbursable contracts pass back to the customer.

The budget backdrop is expansionary. SpaceX files nothing on it, so the account here is Lockheed Martin's, from the U.S. Government Budget Environment heading in Item 2 of its 10-Q for the quarter ended 28 June 2026, accession 0001628280-26-049411, which is also the source of the two contract mix figures above. The FY2026 National Defense Authorization Act authorised $901bn for national defense and the Consolidated Appropriations Act provided $839.2bn of discretionary funding. The FY2027 request seeks $1.5tn including $760bn of weapons procurement and modernisation, with Golden Dome missile defence named as a priority. None of it is appropriated yet.

One customer accounted for 17.9 per cent of first half 2026 revenue and buys from all three segments; a second accounted for 12.2 per cent. Backlog was $28,377m at 31 December 2025, about one and a half times that year's revenue, which is small next to a defence prime's multiyear order book.

What the filing says about competition

The prospectus names competitors readily and ranks itself against none of them with a number. For the Space segment it names United Launch Alliance, Arianespace, Northrop Grumman, Blue Origin, Rocket Lab, Firefly Aerospace and Relativity Space, and claims a meaningful advantage in the breadth of its launch solutions and its cadence. For the Connectivity segment it names terrestrial fixed and mobile operators alongside EchoStar, SES, Telesat, Viasat, Amazon LEO, Blue Origin's TeraWave, Eutelsat OneWeb and Iridium NEXT, and notes that some of those competitors are also its launch customers. For Starlink Mobile it names AST SpaceMobile, Lynk, Globalstar and Skylo. For the AI segment it names OpenAI, Anthropic, Google, Meta and Microsoft, and expects to become a competitor to CoreWeave, Nebius and the hyperscalers as it sells excess compute.

The filing names four limits on its own position. It states that some competitors, particularly in the AI segment, have greater financial, technical or manufacturing resources, and that it has a limited number of customers for its AI products compared with certain competitors. It states that Starlink may be less competitive in dense urban areas where terrestrial fibre and wireless offer higher capacity, lower cost or more consistent performance. It records that while it has historically outperformed certain competitors in the Space and Connectivity segments, there is no assurance it maintains that position. The barriers it claims, capital requirements, spectrum and orbital access, regulatory licences and customer relationships, are the same barriers that protect Iridium and Viasat, and Table 4.4 shows what those barriers have been worth to them.

The Connectivity segment earns a 37.7 per cent operating margin, higher than any listed satellite operator in the set. The other two segments lose $4,930m between them in a half year, against no listed comparator on one side and a defence prime with four times the revenue on the other. No peer group answers whether the Connectivity segment funds the other two long enough for either to become profitable.

5 Material Events & Contracts

In its first sixty three days as a public company SpaceX raised $86.25 billion of equity and $25.0 billion of debt, retired a $20.0 billion bridge loan, refinanced the fixed rate borrowings it had inherited into notes carrying a filed weighted average coupon of 5.855%, and closed a $60.0 billion all stock acquisition. On this report's own principal weighting the three fixed rate instruments those notes replaced averaged 10.70%; section 5.3 gives the weights. Eight 8-K filings cover that period, six of them inside the first fortnight. The xAI merger of 31 January 2026, the EchoStar spectrum purchase of 5 November 2025, the bridge loan of 2 March 2026 and the $5.0 billion revolver of 19 May 2026 were all signed before the prospectus and filed as exhibits to it.

5.1 The listing

Registration statement 333-296070 became effective at 10:00 on 11 June 2026. The company sold 555,555,555 Class A shares to the underwriters at $134.10 against a $135.00 public price and the underwriters took the full 83,333,333 share option, so 638,888,888 shares were sold. Gross proceeds were $86,250 million and net proceeds $85,675 million after $575 million of underwriting commissions and offering costs. The underwriters' discount is $0.90 a share, 0.667% of the offer price. Trading began on Nasdaq and on Nasdaq Texas on 12 June under SPCX; the offering closed on 15 June. Stated use of proceeds: AI compute infrastructure, launch infrastructure and launch vehicles, satellite constellation scale, and general corporate purposes.

Table 5.1 The listing sequence, 30 March to 15 June 2026

5.2 Every 8-K since listing

Table 5.2 Current reports filed since listing, newest first

The 8-K record is thin on operations and heavy on capital structure. Six of the eight reports concern the offering, the notes or the merger. Only one, the 17 June board appointment, is governance, and only one, the 4 August release, is trading.

5.3 The June debt sequence

The three June notes filings are one transaction reported in three stages: launch on 22 June, pricing on 23 June, closing on 26 June. The tranches run from five to thirty years at coupons from 5.350% to 6.650%, a weighted average of 5.855% and a weighted average maturity of 11.7 years. Each tranche is callable at a make whole price to its par call date at the Treasury rate plus 20 to 30 basis points, and at par thereafter. At 30 June the notes carried at $24,852 million against a fair value of $24,697 million.

Table 5.3 The five senior note tranches, issued 26 June 2026

This is the report's single tranche table. Chapters 2, 3, 8 and 9 refer to it rather than restating it.

The bridge loan of 2 March 2026 cleared five instruments inherited with xAI and X, of which three carried fixed coupons: the X B-3 Term Loan at 9.50% on $5,966 million, the xAI Fixed Rate Term Loan at 12.50% on $995 million and the xAI 12.5% Senior Secured Notes on $3,000 million, all as stated in Note 10 of the prospectus. The $995 million is the principal outstanding in Note 10's debt table at 31 December 2025; the facility itself was executed at $1,000 million. Weighted by principal those three average 10.70%, which is this report's calculation and not a figure the filings print. Extinguishing them cost $1,526 million, including a $518 million prepayment penalty on the xAI 12.5% notes. Repaying the bridge in June cost a further $18 million. The $25.0 billion raised exceeded the $20.0 billion bridge by exactly $5.0 billion.

5.4 Material contracts

Fourteen filed exhibits carry the agreements that bind this company: the ones that sold the stock, the ones that bought the two acquisitions and the spectrum, the ones that raised and repaid the debt, and the ones that govern shareholders and employees. Every one was read from the exhibit itself. Table 5.4 lists them; the discussion that follows says what each document does, and quotes the exhibit where the wording carries the term. Two matters of scope follow. The S-1 filed nine Item 601(b)(10) exhibits, EX-10.1 through EX-10.9, of which three are taken below, the indemnification form, the EchoStar purchase and the bridge loan; EX-10.2 through EX-10.7 are not examined here. And the S-8 of 12 June 2026 registers seven plans rather than the two set out below, adding the A&R 2015 plan by reference and the Swarm Technologies, xAI Holdings, xAI Corp. and X Holdings plans inherited with the acquisitions.

Table 5.4 Material contracts, by exhibit

The offering. The underwriting agreement sells 555,555,555 Firm Shares at a purchase price of $134.10 against a public offering price of $135.00, a gross spread of $0.90 a share and $500.0 million on the base deal, with payment and delivery at 6:00 a.m. New York time on 15 June 2026. The option is over up to 83,333,333 Additional Shares at the public offering price, so the banks take no discount on option shares, and the window runs from signing rather than from the prospectus:

The Representatives may exercise this right on behalf of the Underwriters in whole or from time to time in part by giving written notice to the Company not later than 30 days after the date of this Agreement.

The agreement is dated 11 June 2026, so the option ran to 11 July 2026 (8-K 0001628280-26-043288, EX-1.1). Schedule I allocates 111,111,111 shares each to Goldman Sachs and Morgan Stanley, 83,333,333 each to BofA Securities, Citigroup and J.P. Morgan, 11,111,111 each to five more banks and 2,314,815 to each of eleven others, with 2,314,814 to SG Americas Securities. Schedules V to IX carry a Japanese public offering without listing and the Swiss, Australian and UK tranches. The directed share programme runs through six of the banks: Morgan Stanley is the U.S. Directed Share Underwriter and RBC Capital Markets, UBS Securities, Mizuho Securities USA, Macquarie Capital (USA) and ING Financial Markets are the International Directed Share Underwriters. The company lockup in Section 6(o) runs 180 days from the U.S. prospectus date and is releasable only with the prior written consent of Goldman Sachs; the holder lockup in Exhibit A releases in the staged ladder set out in chapter 7. Schedule IV, headed PERSONS OR ENTITIES DELIVERING LOCK-UP AGREEMENTS, is blank in the filed exhibit, so the signatory list is not in the record. Section 10 lets the underwriters walk on a trading suspension, a settlement disruption, a banking moratorium or a material adverse change in financial, political or economic conditions.

The charter, and the two versions of it. The certificate creates Class A at one vote, Class B at ten, Class C with no vote except as required by law, and a preferred class whose terms the board fixes by resolution, all at $0.001 par. The instrument fixes control: “Founder” is defined as Mr. Elon Musk, Class B converts one for one into Class A at the holder's option and automatically on any transfer that is not a Permitted Transfer, and Permitted Transfers run to family members, permitted entities, trusts and nonprofits, so the ten vote stock can move within the family and its vehicles without converting. The vote ratio is one sentence of Article VI (8-K 0001628280-26-043288, EX-3.1):

Each holder of shares of Class B Common Stock shall be entitled to ten (10) votes for each share thereof held.

Two different documents were filed under this exhibit number and they do not authorise the same capital. The version filed with the 8-K of 15 June 2026 reads that the total number of shares the corporation is authorized to issue is “fifty-three billion eight hundred fifty-seven million one hundred fifty thousand (53,857,150,000) shares”, of which 5,325,000,000 are Class B. The version filed as EX-3.1 to the 10-Q of 4 August 2026 reads “fifty-four billion six hundred fifty-seven million one hundred fifty thousand (54,657,150,000) shares”, of which 6,125,000,000 are Class B: the same Class A, Class C and preferred counts, and 800,000,000 more Class B. The 10-Q instrument is the current one and is the basis of Table 6.1, and the 424B4 Description of Capital Stock uses the same 6,125,000,000 Class B figure. Any comparison of the authorised capital across those two filings has to name which instrument it is reading.

The bylaws. Article II carries the meeting, quorum, proxy, record date and advance notice mechanics, and a special meeting route in which a requesting shareholder must first ask the board to fix an ownership record date, which the board may decline to treat as validly requested. Section 9.4 elects into TBOC Section 21.419 and, while the company is listed, sets the ownership threshold for bringing a derivative suit under Section 21.552(a)(3) at “three percent (3%) of the outstanding shares of the corporation.” Section 9.6 binds pre-IPO shares to the 180 day lockup described in the preliminary prospectus of 3 June 2026. Article X sets the dispute regime: Section 10.1 sends internal disputes, including state or federal securities and trade regulation claims and claims against shareholders, controlling persons, directors, officers and underwriters, exclusively to the Texas Business Court, Eleventh Division, bars class, mass and collective actions, and in Section 10.1(c) waives a jury trial in capitals. Anything a court holds to fall outside that court's jurisdiction does not return to the ordinary courts:

Other Disputes shall be exclusively and finally settled by arbitration under the Expedited Procedure Provisions of the Rules (the “Arbitration Rules”) of the International Chamber of Commerce (“ICC”), pursuant to Article 30 thereof

One arbitrator sits on claims up to $5,000,000 and three above it, and again individually only (8-K 0001628280-26-043288, EX-3.2).

Cursor, and the break fee that runs one way. The merger agreement buys Anysphere, Inc. by merging X67 Inc. into it. The price is fixed in dollars and floating in shares:

calculated by dividing (i) an equity value of sixty billion dollars ($60,000,000,000), without any increase for the aggregate exercise price of any Company Options, by (ii) the sum of the Company’s fully diluted capitalization

That per share value is then divided by the Parent Share Price, which for a public SpaceX is the volume weighted average closing price over the seven consecutive trading days ending the day before closing; Schedule 1 works the arithmetic through on an illustrative 10 April 2026 close. It follows a call option agreement of 19 April 2026, support, proxy and contingent release agreements from certain Anysphere holders, and a written stockholder consent promised within 48 hours of signing. The end date is 270 days from signing, extendable by either party for up to 90 days where only the antitrust conditions are outstanding and then by SpaceX alone for a further 90.

Termination fees exist and they run one way. Section 11.04(b) provides a $4,000,000,000 Regulatory Total Fee payable by SpaceX if the deal dies on antitrust with every other condition met. Section 11.04(c) provides a Total Fee payable by SpaceX if Anysphere terminates for SpaceX breach or failure to close:

then Parent shall pay or issue to the Company an aggregate amount equal to ten billion dollars ($10,000,000,000) (the “Total Fee”)

Only one of the two is ever payable, a public SpaceX must pay in cash within fifteen business days, and once paid the fee is the sole and exclusive remedy. Of it, 15% is treated as the termination fee under the merger agreement and 85% as a deferred services payment under a separate Compute Agreement between the parties, which is the arrangement described in section 5.5 below (8-K 0001628280-26-043411, EX-10.1).

xAI, on the opposite terms. The xAI merger is a first merger of K2 Merger Sub Inc. into X.AI Holdings Corp., then a second merger of the survivor into K2 Merger Sub 2 LLC, which survives as X.AI Holdings LLC. Each xAI low vote share converts into 0.1433 of a SpaceX Class A share and each high vote share into 0.1433 of a Class B share, with restricted shares carrying their vesting terms across and share counts rounded up. Holders who are Service Providers could instead elect cash at $75.46 a share, only for shares that were not restricted at the election deadline, through an online system after the first effective time, with failure to elect converting to stock automatically. The xAI board is described as a sole director. The termination terms differ from Cursor's:

This Agreement may be terminated and the Transactions, including the Mergers, may be abandoned at any time prior to the First Effective Time only by the mutual consent of SpaceX and xAI.

There is no outside date, no walk away right and no break fee, and on termination no liability except for fraud (S-1 0001628280-26-036936, EX-2.1).

The spectrum. The EchoStar agreement transfers US rights and licences to 50 MHz across the 2000 to 2020, 2180 to 2200, 1915 to 1920 and 1995 to 2000 MHz ranges, the matching international authorisations and ITU priorities, and up to 15 MHz of AWS-3 spectrum at 1695 to 1710 MHz. The consideration is stated as one number:

The total consideration for the conveyance, transfer, delivery and assignment of the Seller Licenses and Foreign Assets at the Spectrum Acquisition Closing will be $19,616,737,853

It is paid in three layers: first cash to the trust sufficient to discharge EchoStar's high yield notes and release the trust guarantee, then the share leg,

first, up to $11,116,737,853 (as it may be adjusted as set forth in Section 2.1(d), the “Equity Amount”) in Purchaser Shares (rounded to the nearest whole share), valued at a per share price of $212.00

and then any balance in cash. The $212.00 was struck before the five for one stock split of 4 May 2026 and is equivalent to $42.40 after it, and a companion debt service loan agreement has SpaceX funding EchoStar's interest through loans to the trust. Two closings separate the licences from the money: a Spectrum Transfer Closing into the trust, which happened on 22 May 2026, and a Spectrum Acquisition Closing to SpaceX free of liens, expected in November 2027. Either party may terminate if that second closing has not happened by 15 December 2027, extendable to 15 June 2028 on outstanding regulatory conditions and then to 15 December 2028 at SpaceX's option with EchoStar's consent, and SpaceX may terminate if the first closing has not happened by 31 December 2026. One recital of the agreement reads:

WHEREAS, upon the Spectrum Acquisition Closing (as defined below), Purchaser will transfer and deliver to Seller a Starlink satellite for display at a domestic EchoStar location designated by Seller, at no cost to Seller.

(S-1 0001628280-26-036936, EX-10.8.)

The indenture. It issues the five tranches at the sizes and coupons priced on 23 June, pays interest on 15 January and 15 July with record dates on the preceding 1 January and 1 July, and carries almost no restrictions. Section 4.08 is a negative pledge and nothing more:

The Issuer will not (nor will the Issuer permit any of its Domestic Subsidiaries to) create, incur, assume, suffer or guarantee any Secured Debt without making effective provision for securing the Notes equally and ratably with such Secured Debt.

Ten carve outs sit under it, and Section 4.09 adds an exempted secured debt basket of 7.5% of consolidated total assets. There is no maintenance financial covenant, and holders have no right to require repayment on a change of control. Section 3.08 states plainly that “The Issuer is not required to make mandatory redemption or sinking fund payments with respect to the Notes.” Section 3.09 allows a Tax Credit Event redemption at 101% of principal, available if the issuer determines there is a material risk that having sold notes to specified foreign entities within the meaning of Section 7701(a)(51)(B) of the Code would cost it Section 38 tax credits. Notice is available only until the later of the year end of issuance and six months after issuance. Every other redemption right in the document is priced off Treasuries; this one redeems at a fixed 101 per cent of principal. The indenture is signed for SpaceX by Richard Lee, Treasurer, and for the trustee by Ann M. Dolezal, Vice President (8-K 0001628280-26-045763, EX-4.1).

The registration rights agreement. It covers all five tranches, defines the issue date as 26 June 2026, requires the exchange offer to be held open at least 20 business days, and commits the company to “consummate the Exchange Offer on or prior to the 540th day following the Issue Date.” Missing that deadline triggers additional interest, at 0.25% a year for the first 90 days and a maximum of 0.50% a year thereafter. It accrues from the 541st day and stops when the notes are exchanged or cease to be transfer restricted. A shelf registration route carries its own 510 day test. Additional interest is the sole and exclusive remedy, and Section 9 is headed NO UNDERWRITTEN REGISTRATIONS (8-K 0001628280-26-045763, EX-4.7).

The revolver. The agreement provides five billion dollars of unsecured revolving commitments maturing 19 May 2031, extendable lender by lender under Section 2.14, with a $2,000,000,000 performance letter of credit sublimit, a $150,000,000 financial letter of credit sublimit and a $150,000,000 swing line, each part of and not additional to the commitments, and euros and sterling available as alternative currencies. Section 7.06 is the only financial covenant:

Permit the Consolidated Leverage Ratio as of the end of any Measurement Period ending as of the end of any fiscal quarter of the Borrower to be greater than 3.75 to 1.0

The borrower may step that up to 4.25 to 1.0 for the four quarters after a qualified acquisition, but the step up is capped in two ways: at least one quarter back below 3.75 times must follow one before another can be elected, and “there shall be no more than three (3) Leverage Increase Periods during the term of this Agreement.” Material Subsidiary is set at 5.5% of consolidated total assets or revenue. A separate defined term, Material Starlink Asset, captures any Starlink asset or assets that generated revenues above 10% of consolidated revenues in the most recent measurement period, or that are otherwise material to the operation of Starlink (S-1/A 0001628280-26-039276, EX-10.10). Nothing was drawn in the half year to 30 June 2026. Pricing turns on a ratings grid, set out in full below.

Table 5.5 Revolving credit facility applicable rate, by debt rating

The bridge. The purpose is stated on the face of the agreement, to refinance the existing subsidiary indebtedness, and the size in its first recital:

The Borrower has requested that the Lenders extend credit to it in the form of a term loan credit facility in an aggregate principal amount of $20,000,000,000 (the “Bridge Facility”).

It matures 2 September 2027, extendable by the borrower twice for three months each on ten to thirty days' notice against an extension fee of 0.25% of the extended principal, with duration fees of 0.125% of loans outstanding at one year and 0.25% at fifteen months. Its applicable rate grid runs to six levels against the revolver's five, topping out at 1.750% over Term SOFR below BB+/Ba1/BB+. The second funding date guarantors are X Corp., a Nevada corporation, X.AI LLC, a Nevada limited liability company, and CTC Property LLC, a Wyoming limited liability company. Section 2.05(b) sets the mandatory prepayment terms: 100% of the net cash proceeds of any debt issuance within three business days, and, on a qualified IPO,

then the Borrower shall prepay the Loans in an amount equal to 100% of such Net Cash Proceeds not later than the date that is six (6) months following the date on which the Borrower or such Subsidiary receives such Net Cash Proceeds.

Voluntary prepayment carries no premium or penalty, and the facility was repaid in full in June 2026 (S-1 0001628280-26-036936, EX-10.9).

The shareholder and employee agreements. The investors' rights agreement gives the Series A to Series N holders demand and piggyback registration rights, with the company paying registration expenses other than underwriting discounts and the fees of counsel for the selling holders, and information rights for major investors other than competitors. The rights are self extinguishing: no holder may exercise them “after the earlier of (i) two (2) years following the consummation of a Qualified IPO”, the point at which Rule 144 permits a full sale within three months, a holding below 1% of outstanding capital stock, or a change of control. The exhibits listing the Series A to Series N investors are not part of the filed document and no investor signature page is filed, so the agreement names its parties only in the body. The form of indemnification agreement contracts advancement of expenses before final disposition and indemnification to the fullest extent the TBOC permits, with the procedure tilted toward the director: “Upon making such request for indemnification, the Indemnitee shall be presumed to be entitled to indemnification and to be held harmless hereunder and the Company shall have the burden of proof in making any determination contrary to such presumption.” A determination not made within 60 calendar days is deemed favourable, and after a change in control the company must indemnify settlements approved by independent counsel.

The two employee plans, and what they actually carry. The A&R 2024 Equity Incentive Plan was adopted by the board on 11 December 2024, approved by stockholders on 21 December 2024, restated by the board on 10 June 2026 and terminates on 10 December 2034. Its reserve is stated as “the aggregate number of shares of Common Stock that are available to be issued pursuant to Equity Awards from and after the Amendment Date is 300,894,150 shares of Common Stock”, already adjusted for the Class C reclassification and the five for one split of 4 May 2026. It does not absorb the awards outstanding under the earlier plans: those remain subject to the terms of the 2015 and 2012 plans, and only shares that expire, are forfeited, are repurchased at original issue price or are withheld for tax flow into the new reserve. At the amendment date 20,906,545 shares had arrived that way and 112,979,230 shares were still subject to outstanding 2015 Plan awards, which is the overhang that sits alongside the 300,894,150 rather than inside it. No new awards may be granted under the 2015 Plan, there is no automatic single trigger vesting on a change in control, and participation is conditioned on binding arbitration under an employee arbitration and class action waiver agreement. The employee stock purchase plan has two share numbers and only one of them is the plan's capacity: the aggregate plan cap is 75,000,000 shares, and “the aggregate number of shares of Common Stock that are available to be issued under the Plan from and after the Amendment Date is 24,026,920 shares of Common Stock.” Purchases are made at not less than the lesser of 85% of fair market value on the offering date and 85% on the purchase date, and offerings may not exceed 27 months. A reset applies whenever fair market value on the first day of a new purchase period is at or below the value on the offering date: the offering terminates and participants are enrolled in a new one (S-8 0001628280-26-042832, EX-99.1 and EX-99.2).

Quotations in this section are reproduced exactly as filed, including each exhibit's own date forms, capitalisation and punctuation.

5.5 Agreements described but not filed

Five commercial arrangements carry economics disclosed in the filings without an exhibit behind them. Note 20 of the 10-Q, Acquisitions, carries two of them: the Cursor call option agreement of 19 April 2026, and the concurrent Cursor compute agreement, under which SpaceX supplies GPU cluster capacity and the two collaborate on models including Grok. The option was carried at an initial fair value of zero, and the compute arrangement was not material in the second quarter. The Mesh Optical merger agreement of May 2026, roughly 3.8 million Class A shares with a cash election of up to $2.5 million a holder for Class B, closed on 6 July 2026 and appears in the same note, its equity plan filed as Exhibit 10.4 to the 10-Q. The $14.1 billion of cloud services agreements and the Starshield awards above $6 billion are quantified in the 4 August release without naming a counterparty there, though the prospectus notes disclose the Anthropic cloud services agreement and its 90 day termination right. Against those the balance sheet carries $27,955 million of noncancellable contractual commitments at 30 June, $22,244 million of it falling in 2027, plus $645 million of letters of credit and $465 million of surety bonds.

Source: SEC filings retrieved through the SEC-API.io MCP server. CIK 1181412.

6 Ownership, Voting Power & Annual Meeting

SPCX is a controlled company. Elon Musk owns about 46 per cent of the equity and casts about 85 per cent of the vote. Class A, the only listed class, carries 58.4 per cent of the economics and 12.3 per cent of the vote on the 28 July 2026 share counts. No annual meeting has been held and no proxy statement exists; the bylaws treat the 2026 meeting as already held and point the next one at June 2027.

Two classes, one listed, ten votes to one

The charter that took effect at the offering creates three classes of common stock and one blank cheque preferred. Class A carries one vote, Class B carries ten, and Class C carries none. Class C was reclassified into 494,050,675 Class A shares before the offering and none is outstanding; the 10.00bn authorised sit unissued alongside 2.40bn authorised preferred. The two Form 8-A registrations filed on 10 June both register the same Class A common stock under the same Exchange Act file number, one on The Nasdaq Stock Market LLC and one on Nasdaq Texas, LLC. Class B is not registered and does not trade.

Table 6.1 Capital structure and voting rights, share counts at 28 July 2026

Authorised counts and vote ratios are as filed in Article VI(A) of the Restated Certificate of Formation, EX-3.1 to the 10-Q of 4 August 2026, accession 0001628280-26-052535, and as described in Description of Capital Stock in the 424B4 of 12 June 2026, accession 0001628280-26-042639. Outstanding counts from the 10-Q cover page. This table is the report's single statement of the authorised capital, and the instrument behind it is the one described in section 5.4. Two instruments were filed under the same exhibit number and they differ: the version filed with the 8-K of 15 June 2026, accession 0001628280-26-043288, authorises 53,857,150,000 shares in total with 5,325,000,000 of Class B, while the version filed with the 10-Q authorises 54,657,150,000 with 6,125,000,000 of Class B, an increase of 800,000,000 Class B shares. The 10-Q version is the current one, it is the one the prospectus describes, and it is the one used above.

The founder's position

The prospectus cover puts Musk at 82.4 per cent of the combined voting power, of which 81.1 points came from Class B alone, or approximately 82.3 per cent if the underwriters exercised their option to purchase additional shares in full. They did, and the 8-K of 15 June 2026 reports the offering closing at 638,888,888 shares (0001628280-26-043288), so 82.3 per cent is the figure that applies to the offering as executed. Six weeks later his voting share had risen without a single share changing hands. Class B converts one for one into Class A on almost any transfer, so every other holder who sold shrank the ten vote class and enlarged the one vote class. Between the offering and 28 July, 210,181,989 Class B shares converted away and Class A grew by 316,096,759. His share of Class B rose from 91.6 per cent to 95.1 per cent, and his voting power on the 28 July share counts rises to 84.8 per cent against 46.0 per cent of the equity. The charter sets no limit on that increase. The prospectus description of capital stock sets no time based and no ownership based sunset: the Class B right to elect a majority of the board holds for so long as one Class B share remains outstanding.

Table 6.2 Economics against votes, the founder's position on the filings' own bases

Holdings from the 424B4 cover page and Prospectus Summary. The offering column is struck on the share counts with the underwriters' option exercised in full, the basis on which the deal was executed, which is the 82.3 per cent the cover gives for that case; the cover's 82.4 per cent is the base offering without the option. The 28 July column applies the same holdings, confirmed unchanged at 30 June by the 13G at accession 0001104659-26-095936, to the 10-Q cover page share counts. The prospectus also prints 84.4 per cent in its beneficial ownership table; that figure deems his option shares outstanding and uses a 1 May share base.

His holding includes 1,302,072,285 restricted Class B shares whose vesting turns on market capitalisation and operational milestones. The prospectus states that he may vote them now; the milestones condition the ownership of those shares.

The 5 per cent register

Four Schedule 13G statements have been filed, all captioned to Class A, all with an event date of 30 June 2026 and all under Rule 13d-1(d), the exempt investor route used by holders who acquired their stock before the company registered under Section 12.

Table 6.3 Schedule 13G filings, all captioned to Class A common stock

Every one of the four is captioned to Class A, and three of the four report only Class A. Musk's shares are stated on an as converted basis and include 5,569,053,075 Class B; his voting power is measured with those option shares deemed outstanding. Item 4 of the Thiel filing shows that its 427,306,025 total is not all Class A either: 3,748,835 of it is Class B held directly by Barrel Vault LLC, convertible one for one into Class A. That is 0.9 per cent of the position and 0.07 per cent of the Class B outstanding, so it does not move the voting column. Percentages of Class A are as filed and reproduce against the 10-Q cover count.

The three holders other than the founder own 1.48bn shares between them on the as converted basis their filings use, 19.3 per cent of the listed class, and 2.4 per cent of the vote.

The institutional register after one quarter

The June quarter is the first for which SPCX could appear in a Form 13F. The stock appears in 1,815 of them, counted as every Form 13F holdings report on the SEC-API.io index carrying CUSIP 84615Q103, with no restriction on the period reported. The 1,795 reports aggregated below cover 1,762 managers holding 3.65bn shares worth $624bn, 47.4 per cent of Class A. Read against the vote rather than the equity, the same holdings come to 5.8 per cent.

Table 6.4 The twenty largest reported 13F positions, quarter ended 30 June 2026

Long common stock only; put and call positions are excluded. Where a manager amended its report, the latest report is used. Aggregated from 1,795 of the 1,815 reports that carry CUSIP 84615Q103, that being every Form 13F holdings report on the SEC-API.io index carrying the CUSIP, with no restriction on the period reported.

Only two of the twenty are index complexes. The Vanguard Group Inc and State Street Corp report no position in the CUSIP at all, and Vanguard appears only through subsidiary managers. Alphabet, Valor, FMR, VY Capital, Gigafund, the Public Investment Fund, SC US, a16z and Founders Fund, all of them holders from the private rounds, occupy the places index money usually holds. The rest of the register is numerous and small. Of 1,762 managers, 1,491 hold under 100,000 shares each and 0.4 per cent of the reported total between them, while 39 managers hold 91.4 per cent.

Who the filings name, and who they do not

The private rounds that built this register named nobody. Form D has no field for a purchaser's identity, as chapter 1 sets out, so the thirteen years and $9,034m behind Table 1.6 come with an investor list of exactly zero names. Every investor name in the record comes instead from one of four documents: the prospectus beneficial ownership table with its footnotes, the prospectus section on certain relationships and related person transactions, the Amended and Restated Investors' Rights Agreement filed as exhibit 4.2 to the S-1 of 20 May 2026, and the four Schedules 13G.

Table 6.5 Investors and holders the filing record names

Three of those thirteen carry no size at all. Fidelity and Baillie Gifford & Co. appear only in the body of the investors' rights agreement, in the definitions and in a clause permitting the two of them to disclose publicly what they had invested; DFJ Growth appears as an investor only in the prospectus sentence naming entities affiliated with Elon Musk, Google, Valor and DFJ Growth as parties to that agreement, and otherwise only as the firm Randy Glein co-founded, in his director biography. No filing by the company states a number for any of the three, and this report does not supply one. FMR LLC and Baillie Gifford & Co report positions of their own in Table 6.4, but a Form 13F states only what the manager held at a quarter end.

The two disclosure regimes give different answers. Under the heading 5% Shareholders the prospectus table lists exactly one name, Elon Musk. Google at 7.2 per cent of Class A and Thiel at 5.5 per cent appear nowhere in that table; both surface only in the 13Gs of 14 August, two months after the listing. Gracias is in the table, but under directors rather than under the 5 per cent heading. The prospectus does disclose the Google relationship elsewhere, as a party to the investors' rights agreement dated 4 August 2020 and through the biography of director Donald Harrison, but a reader of the ownership table alone would not have known that a 7 per cent holder existed.

Two lists that would have closed the gap are absent from the filed exhibits. Exhibits A to N to the investors' rights agreement are the Series A to Series N investor schedules, and they are not part of the filed document; no investor signature page is filed with it either. Schedule IV to the underwriting agreement, headed as the list of persons or entities delivering lock up agreements, is blank as filed. The register that chapter 7 describes as covering roughly 12.5 billion shares therefore has no named signatories in the record.

No annual meeting, and what stands in its place

No annual meeting of shareholders has been held and no DEF 14A has been filed. Two dates in the bylaws filed as exhibit 3.2 to the 8-K of 15 June (0001628280-26-043288) supply a timetable of the company's own making. The 2026 annual meeting is deemed to have been held on 19 June 2026 for the purpose of the special meeting restrictions. Shareholder nominations and proposals for the next annual meeting must be delivered between 120 and 90 days before the first anniversary of the preceding meeting, which puts that window between 19 February and 21 March 2027. On the bylaws' own arithmetic the next meeting is anchored to 19 June 2027, and a shareholder who wants a name or a proposal on its agenda has to move in the first quarter of that year. No filing in the record fixes a date for the meeting itself.

Table 6.6 The first annual meeting timetable as the governing documents set it

Every date in the table is taken from the charter and the bylaws filed as exhibits 3.1 and 3.2 to the 8-K of 15 June 2026, accession 0001628280-26-043288, and to the 10-Q of 4 August 2026, accession 0001628280-26-052535. No filed document sets a latest date by which the first annual meeting must be held.

The absence of a meeting matters less than it would at a widely held company, because the charter routes most decisions around meetings altogether. Shareholders may act by written consent at the minimum vote that would carry the matter at a meeting, and the founder holds more than that minimum on his own. Article XIII of the charter lets the board, the chairman, the chief executive, the Founder and, to the extent the Texas Business Organizations Code requires, the president call a special meeting. Shareholders may call one only at 50 per cent of the outstanding stock entitled to vote, a threshold no combination of outside holders can reach. The bylaws then adopt the Texas Business Organizations Code section 21.373 test for shareholder proposals: $1m of stock or 3 per cent of the voting shares, held six months, plus solicitation of holders representing 67 per cent of the voting power. The founder alone holds more than 67 per cent, so that last condition cannot be met without him.

Table 6.7 The control provisions and what each one does

The company takes the controlled company exemptions from the Nasdaq and Nasdaq Texas standards on board and committee independence, keeping only the fully independent audit committee that no exemption covers. The prospectus put the board at eight directors on completion of the offering, five of them Class B Directors. The 8-K of 17 June 2026 then reports that on 16 June the board elected Roelof Botha an independent Common Stock Director to fill the existing vacancy and appointed him to the audit committee (0001628280-26-043865), and he filed a Form 3 on 18 June (0001628280-26-044455). The board is therefore nine seated directors: five Class B Directors, and four Common Stock Directors in Ira Ehrenpreis, Randy Glein, Steve Jurvetson and Botha, whom Class A and Class B elect together. Fifty one per cent of nine, rounded up as the charter requires, is still five. On the numbers in Table 6.2 the founder carries the Common Stock Director vote as well.

How the register was built, and what the public paid for it

The concentration is thirteen years of private financing arriving intact, set out offering by offering in Table 1.6: fifteen offerings between March 2009 and August 2022 that sold $9.03bn, all under Rule 506.

The prospectus dilution table sets out the trade plainly. New investors bought 4.2 per cent of the shares for 48.0 per cent of the total consideration ever paid into the company, at $135.00 against an average of $6.48 for existing holders, and took an immediate $127.15 per share of dilution to adjusted pro forma net tangible book value of $7.85.

New money supplied roughly half the capital ever paid in, took 4.2 per cent of the shares, and holds one eighth of the vote. Nothing in the charter, the bylaws or the shareholder filings made so far changes those proportions before 2027, and the automatic conversion of Class B moves them further the other way every time a Class B holder sells.

7 Insider Activity

There is no insider trading record at SPCX to read. In the 85 days from the first trade on 12 June 2026 to 5 September 2026, insiders filed ten Forms 3 and one Form 4, and not one of those filings reports an open market purchase or sale of SPCX stock. A search of Forms 3, 4, 5 and 144 on CIK 1181412 through 5 September 2026 returns eleven documents in total and nothing else: no Form 5, no Form 144, and no Form 4 after 17 June 2026. The buy to sell ratio in this chapter is computed on open market transactions only, and there are none, so the ratio does not exist. The subject of the chapter is therefore the lockup, which restricts every one of the roughly 12.5 billion shares that existed before the offering and releases them in seventeen steps running to the third quarter of 2027.

7.1 The Forms 3 are a roster

Nine Forms 3 were filed on 11 June 2026, the evening the registration statement went effective. A tenth followed on 18 June for Roelof Botha, whom the board elected as an independent director and audit committee member on 16 June 2026 (8-K 0001628280-26-043865). Nine names is exactly the count of executive officers and directors the prospectus expected to serve on completion of the offering, so the Section 16 roster is complete.

A Form 3 states what a person already owned on the day the class registered. It records no transaction and no decision. Ten of them filed on listing day identify who the insiders are. They carry no information about whether anyone is buying or selling.

Table 7.1 The Section 16 roster and its holdings at listing, from the Forms 3

Four features of the roster matter. First, Musk holds 68% of the Class A equivalent common stock the ten forms report, and that is before the 1,302,072,285 unvested restricted Class B shares his form names in its remarks but excludes from its tables. Of those, 1,000,000,000 sit under the SpaceX CEO Award, which vests across fifteen market capitalisation tranches from $500 billion to $7.5 trillion and on a permanent Mars colony of at least one million inhabitants. The other 302,072,285 sit under the AI CEO Award, which vests across twelve tranches from $1.065 trillion to $6.565 trillion and on completion of non Earth based data centres delivering 100 terawatts of compute a year. Second, Gracias reports his 503.4 million Class A shares as held of record by thirty named Valor entities and disclaims beneficial ownership beyond his pecuniary interest, so the figure measures the funds he is associated with rather than a personal stake. Third, three of the ten report no securities at all. Fourth, only three of the ten are officers; the other seven are outside directors, and the 8-K that appointed Botha states that non employee directors currently receive no cash or equity compensation for board or committee service. That last point removes the usual source of insider filings at a newly listed company: with no director grants there is nothing to vest and nothing to withhold shares against, so the routine Forms 4 that normally start within months of an IPO have nothing to report here.

7.2 What the single Form 4 reports

Form 4 0001628280-26-044069 was filed on 17 June 2026 with a period of report of 2 February 2026. A Form 4 carries the date of the earliest transaction it reports, and this one reports thirteen lines spanning 2 February to 15 June 2026 in a single document. Every line is either acquisition consideration, an award restructuring, an estate transfer or the mechanical conversion of preferred stock at the closing of the offering. The one line coded S is dated 2 April 2026, ten weeks before SPCX traded, when the company was still private, so it cannot have been an open market sale whatever the code.

Table 7.2 Every line of Form 4 0001628280-26-044069, Elon Musk

Share counts on the February to April lines are restated for the five for one forward stock split the company effected on 4 May 2026. The June conversions carry the closing date of the offering, 15 June 2026. Filed two days later, the form is timely against those June events, and the earlier lines ride along because no Section 16 obligation existed before the class registered.

The form leaves Musk holding 849,494,440 Class A shares and 3,916,980,790 Class B shares, plus 350,000,000 Class B shares under a fully vested option at $8.3998 expiring 11 February 2031, and the 1,302,072,285 unvested restricted Class B shares the form excludes. Those four numbers add to 6,418,547,515, which is the up to 6.4 billion shares the prospectus assigns to the founder lockup bucket. The bucket is his entire stake, vested and unvested.

7.3 The lockup: 12.5 billion shares, three pools, seventeen release dates

The prospectus is dated 11 June 2026 and every deadline runs from that date, one day before the first trade. Consent to release rests with Goldman Sachs & Co. LLC on behalf of the underwriters, which may waive at any time. Three pools sit under the agreements (424B4 0001628280-26-042639, Shares Eligible for Future Sale and Underwriting):

• The 180 day pool, about 4,691 million Class A shares. Runs to 8 December 2026, the 180th day. Ten automatic early releases, listed below.
• The extended pool, about 1,760 million shares. Certain shareholders, restricted until the second full trading day after the release of results for the quarter ending 30 June 2027. Six automatic early releases, none before the first quarter of 2027.
• The founder pool, up to 6,400 million shares. All of Musk's stock, restricted to the 366th day, 12 June 2027, with no early release provisions of any kind.

The pool sizes are not printed as totals. They are recovered from the tranche percentages the prospectus does print. The extended and founder pools less the 350 million option shares come to 7,810 million shares, the 7.8 billion the prospectus states for the restrictions lasting over a year, and 59.7% of the 13,076 million shares outstanding after the offering, which is the approximately 60% the prospectus states. All three pools less the option shares come to 12,501 million shares against the 12,520 million outstanding immediately before the offering. Every share that existed before the IPO is locked. The only unrestricted stock at listing was the 638,888,888 shares sold in the offering, 8.3% of the 7,696,293,669 Class A shares on the 10-Q cover at 28 July 2026.

Table 7.3 The release ladder, in prospectus order

Three tranches are resolved, two of them by release and one by failing its price test, and fourteen are ahead. The first two hinged on the first earnings release, which the company furnished on 8-K 0001628280-26-052515 at 16:01 Eastern on 4 August 2026, after the close, making the second full trading day after it 6 August 2026. The 20% tranche released on that date without condition. The additional 10%, 455.8 million shares, required the closing price to reach $175.50, thirty per cent above the $135.00 offering price, on five of the ten trading days ended on and including 4 August 2026.

That test failed. Across those ten sessions the highest close was $125.33, on 4 August itself, so none of the ten reached $175.50 against the five required. The outcome is established from the market record rather than from a filing, because no filing reports it: no SEC filing carries daily closing prices, the company has reported no release, and no Section 16 or Rule 144 filing turns on it. The 455.8 million Additional Release Shares therefore stayed inside the 180 day pool and now fall due on 8 December 2026, taking that date from 328.4 million shares to 797.6 million. Section 9.4 carries the same figure. The 59.1 million affiliate catch up on day 91 implies 44% of the affiliate block released by then, which is 20 plus 10 plus 7 plus 7. The table was therefore drawn on the assumption that the conditional 10% would release.

Cumulative supply released from the 180 day pool by 5 September 2026 is 1,230.5 million shares, 26% of that pool, and that figure is now final. It is 1.9 times the 638,888,888 shares that were freely tradable after the IPO. Between 9 September and 8 December a further 3,461 million shares reach the market across seven dated or results linked steps, so the pool is fully released by 8 December and the tradable Class A count multiplies several times over. The extended and founder pools then hold about 8.2 billion shares off the market for a further six to eight months.

Carve outs exist but they are narrow and mostly do not put stock on the market. Gifts, estate planning transfers, transfers to family members and to controlled entities, transfers by operation of law, distributions by funds to their partners, and transfers to a pledgee under a financing arrangement in place at the offering are all permitted, and for the gift, estate, family, controlled entity and pledge transfers the recipient must sign a lockup for the balance of the term. No Exchange Act filing reporting a reduction in beneficial ownership may be made voluntarily, and where one is legally required its footnotes must state the circumstances of the transfer. Sales to cover exercise price or withholding tax on equity awards are permitted, as are transfers to the company on vesting, settlement, repurchase or a board approved tender offer, and tenders into a board approved change of control. Rule 10b5-1 plans may be adopted during the period, but nothing may be transferred under them until the applicable lockup expires. The company itself is separately restricted for 180 days from issuing or registering stock, subject to exceptions for the offering, existing awards, new plan grants, Form S-8 filings, withholding sales and acquisition consideration. That last exception is live: on 14 August 2026 the company reported both the completion of an acquisition and an unregistered sale of equity securities under Item 3.02 (8-K 0001628280-26-056945). Musk's shares carry none of the early release provisions and only the Permitted Transfers his charter defines.

Two mechanisms sit outside the lockup and enlarge the supply picture. Holders of about 9.2 billion Class A equivalent shares hold registration rights under the investors' rights agreement, and a demand exercised after a lockup lapses converts restricted stock into freely resaleable stock on effectiveness. The prospectus is not consistent on that count: the corporate structure risk factor gives approximately 9.2 billion shares subject to registration rights and approximately 12.2 billion available under Rule 144 from 90 days after the offering, while the Registration Rights section under Shares Eligible for Future Sale puts approximately 12.2 billion under the registration rights themselves. The lower figure is used here. Separately, affiliates selling under Rule 144 face a volume cap of the greater of 1% of Class A outstanding, about 77.0 million shares on the 7,696,293,669 shares on the 10-Q cover, or the four week average weekly volume, in any three month period, and each such sale requires a Form 144. None has been filed to date.

7.4 The insider record to date

No conclusion about insider sentiment can be drawn from what has been filed. The ten Forms 3 are a census taken on day one. The single Form 4 reports a corporate acquisition, an award restructuring, a private transfer and the preferred conversion that the offering itself compelled. None of those lines is a purchase or a sale of listed stock. The chart of net open market activity by month that this chapter would normally carry cannot be drawn, because twelve months of listed history do not exist and the open market transaction count over the weeks that do exist is zero.

The filings give a supply timetable. The dated tranches of the 180 day pool are fixed and unconditional, the results linked tranches move only with the reporting calendar, and the 8 December 2026 expiry is 94 days ahead of 5 September 2026. The first informative insider filings will be the Forms 4 and 144 that follow those releases. Until one appears, the filed record shows no insider open market activity.

8 Risk Factors

Four risks in this filing carry numbers that move the model: Starship launch cadence, launch and in orbit losses that the company states it does not insure and does not expect to insure (424B4 0001628280-26-042639, Item 1A), a fifth of revenue depending on US federal agencies that can cancel contracts for convenience, and $38,433 million of principal debt under covenants that were breached once and waived once in 2026. The remaining risk factors are either general to space and AI issuers or governance provisions bearing on what a shareholder can do. The filing does not rank them that way. It leads with Starship and then writes most about AI regulation, its own bylaws and intellectual property.

This chapter carries no change map, because only one Item 1A has been filed. SpaceX listed on 12 June 2026. It has filed one registration statement, whose final prospectus is the 424B4 of 12 June 2026, accession 0001628280-26-042639, and one quarterly report, the 10-Q for the quarter ended 30 June 2026, accession 0001628280-26-052535. There is no 10-K, so there is no prior year Item 1A, no newly added risk, no reworded risk and no dropped risk. What follows groups the disclosure the way the filing groups it and ranks within groups by the weight the filing itself gives each risk: position, length, specificity and whether a number is attached.

8.1 How the section is weighted

On this report's own count of the section, it runs 29,433 words across 54 individual risk factors under two group headings. Risks Related to Our Business takes 44 risks and 23,490 words, 79.8% of the section. Risks Related to Our Corporate Structure, Ownership of our Class A Common Stock and This Offering takes 10 risks and 5,943 words. The filing states no word count of its own, and a different counting convention for headings and page furniture moves the totals by a few hundred words.

Ten risk factors account for 11,370 words, 38.6% of the section on the same count.

Table 8.1 The ten risk factors the filing writes most about

The Words column is this report's measurement of the filed text; the filing states no word count.

Position and length disagree. Starship opens the section and three risks are longer, all three of them legal. The single longest item in the corporate group, and the second longest in the whole section, describes the company's own bylaw provisions for where and how a shareholder may sue it.

8.2 The risks specific to this company

Table 8.2 The risks that bear on the numbers, ranked

Starship is a single point of dependence, and the filing says so plainly. Falcon 9 and Falcon Heavy cannot carry V3 satellites or V2 Mobile satellites. Next generation Starlink capacity, global satellite to mobile service and orbital AI compute therefore all sit behind one development programme whose flight tests have already drawn FAA mishap determinations, including on the twelfth flight. The filing states that AI compute satellites at scale need full Starship reusability to be economically compelling, and that without it the other programmes proceed more slowly and at higher cost. That statement bears on capital efficiency as well as on schedule.

The insurance disclosure carries a direct balance sheet consequence. The filing states that the company does not typically obtain insurance coverage for its satellites, payloads or launch vehicles, and that as a result it bears the full financial cost of any such losses. It repeats the point on in orbit assets, which it currently does not insure and does not expect to insure in the future. It also states it does not maintain key person life insurance on Mr Musk. A launch failure or a constellation loss therefore lands entirely in the income statement through accelerated depreciation and impairment, with no recovery line.

Government concentration is quantified, and a second disclosure of the same size sits beside it. Item 1A states that approximately one fifth of 2025 revenue came from agencies within the US federal government. On 2025 revenue of $18,674 million that is roughly $3.7 billion. The concentration note in the financial statements separately discloses an unnamed single customer at 20.9% of consolidated revenue in 2025, 24.2% in 2024 and 25.2% in 2023, spanning all three segments, with no other customer above 10%. The note does not identify that customer, so the two figures are equal in size without being formally the same disclosure. Backlog stood at $27,621 million at 31 March 2026, of which $13,236 million was already deferred revenue. The filing then states that government contracts are susceptible to unilateral termination, reduction in scope or delay at the government's convenience, and that Space segment revenue is primarily fixed price, so the company absorbs cost overruns rather than recovering them. Concentration, cancellation rights and fixed price exposure sit on the same revenue base.

Spectrum is the licensing risk with a date and a price on it. The Starlink Mobile plan depends on the EchoStar AWS-4, H block and AWS-3 licences, whose approximately $19.6 billion of consideration is set out in Table 5.4. The FCC approved the transaction on 12 May 2026; closing is expected in November 2027. The filing is explicit that even after closing the licences may not be sufficient, that global rights must still be obtained from national regulators, and that the 5G NR-NTN bands for these frequencies are not supported by radio frequency front end hardware in any commercially available handset. Separately, the Business section states that FCC spectrum licences and authorisations typically run 10 to 15 years and are then subject to renewal. Gen1 carrier partnerships exist in approximately 30 countries, but the filing distinguishes commercial partnerships from the regulatory approvals that permit service.

The debt risk factor is short and the covenants are not in it. Item 1A gives 242 words to indebtedness and one number, $29,132 million of principal at 31 March 2026. The terms sit elsewhere in the filings, and Table 8.3 gathers them.

Table 8.3 What the filings quantify around the leverage and control risks

Every figure in the second column is a disclosed one, taken from the 424B4 (0001628280-26-042639) and the 10-Q (0001628280-26-052535). One consequence of the derivative action threshold is worth adding and is this report's own arithmetic rather than a filed figure: 3 per cent of the post offering share count, valued at the $135.00 offer price, is roughly $53 billion of stock, so the gate is set at a level no ordinary holder reaches.

Neither the risk factors nor the rest of the 424B4 contains going concern or substantial doubt language, and nor does the management discussion in the 10-Q, which states instead that the company believes it has sufficient sources of funding to meet its business requirements for at least the next twelve months.

8.3 What has already partly happened

Table 8.4 Risks partly realised between listing on 12 June 2026 and 5 September 2026

8.4 Part II Item 1A of the 10-Q

The 10-Q carries its own Item 1A, and it supplements the prospectus. The filing says the risk factor set out below supplements the risk factors disclosed under the section titled Risk Factors in the Prospectus, and that except as set out below there have been no material changes.

One risk factor is added, on AI infrastructure and data centre operations. Three elements of it are new relative to the prospectus. First, the company states that an increasing portion of consolidated revenue comes from the AI segment, driven by AI Infrastructure. Second, it states that a significant portion of AI infrastructure revenue is concentrated in a small number of customers. Third, it describes the contract shape: cloud services agreements generally provide for monthly fees and, after an initial period covering capacity ramp, may be terminated by either party on 90 days' notice.

AI segment revenue grew $1,824 million year on year in the quarter and was the largest single contributor to 91.9% consolidated revenue growth, and the company has now disclosed that the contracts behind it are concentrated, monthly and cancellable at 90 days. The prospectus disclosed the Anthropic cloud services agreement and its 90 day termination right in the notes; the 10-Q, seven weeks after listing, raised the termination right to the level of a risk factor.

Source: SEC filings retrieved through the SEC-API.io MCP server.

9 Stock Price, Scenarios & Sensitivity

Three things about SPCX are knowable today. The company raised $110.7bn in the fortnight to 26 June 2026 and put 82.7 per cent of its first half capital expenditure into the AI segment, against 9.5 per cent into Connectivity and 7.8 per cent into Space. Between 9 September and 8 December 2026 the lockup releases 3,461 million further Class A shares on dates already fixed in the prospectus, 5.4 times the number sold in the offering and about 50 days of recent trading volume. And a three year projection built from the drivers the filings report lands anywhere between $19 and $306 per share, with the terminal multiple moving that answer nearly twice as far as the largest operating driver.

9.1 Price basis

The price block below is fixed by chapter 4 and is the single price basis for the whole report. It is reproduced here. The comparator half of the same block sits in Table 4.5 and is not repeated. Nothing in this chapter states a price of its own.

As at the close of trading, Thursday 3 September 2026.

The report date is Saturday 5 September 2026. The 3 September close is used rather than the 4 September close because 3 September is the most recent session for which a settled closing price is confirmed for every one of the eight names in this block on a single, common date. Mixing a 4 September close for some names with a 3 September close for others would put the comparator valuations on different days.

Share counts are the counts on the cover page of each company's most recent quarterly report, which is the count the registrant itself certifies. Market value is the closing price multiplied by the total of all classes of common stock outstanding on that cover page.

Prices are last sale prices from public market data. Share counts and their sources come from SEC filings retrieved through the SEC-API.io MCP server.

Space Exploration Technologies Corp. (SPCX)

Item	Value
Price date	3 September 2026, close
Last price	$149.74
Shares outstanding	13,181,779,945 (Class A 7,696,293,669 + Class B 5,485,486,276)
Share count source	Cover page, Form 10-Q for the quarter ended 30 June 2026, accession 0001628280-26-052535, count as at 28 July 2026
Market value	$1,973.84bn

SPCX market value covers both share classes. Class B carries ten votes per share and does not trade. Treating market value as the value of the traded Class A shares alone would understate it by roughly 42 per cent. The $1,973.84bn figure values all shares at the Class A price, which is the convention used throughout this report.

SPCX's IPO priced at $135.00 per share on 11 June 2026 for 555,555,555 Class A shares, raising $75.0bn before expenses (424B4 cover page, accession 0001628280-26-042639, which is dated 11 June 2026; the registration statement became effective that day, per EFFECT 9999999995-26-001968). Trading began on 12 June 2026. The 3 September price of $149.74 is 10.9 per cent above the offer price.

Share count for the per share arithmetic in this chapter is the same cover page count, 13,181,779,945, struck at 28 July 2026. Two issuances sit outside it and both are filed. The Cursor merger completed on 14 August 2026: Cursor common and preferred stock converted into 389,289,254 Class A shares and vested Cursor restricted stock units into a further 1,752,426, struck on an implied Cursor equity value of $60.0 billion and a SPCX price equal to the volume weighted average closing price over the seven consecutive trading days immediately preceding the closing (8-K 0001628280-26-056945). That is 391,041,680 shares issued. The same filing reports approximately 29,128,326 restricted stock units and approximately 44,365,047 options over Class A stock assumed on unvested Cursor awards; those are carried by the annual dilution assumption in Table 9.3 rather than added here. To the cover page count the scenarios therefore add the 391,041,680 Cursor shares and the 261.8 million shares of consideration due at the Spectrum acquisition closing (10-Q 0001628280-26-052535).

9.2 Fifty eight sessions of price record

The record runs 58 sessions, from the listing to the price date of the block above. The stock closed its first day 19.2% above the $135.00 offering price, peaked at $201.80 on 16 June, 49.5% above the offering price, and troughed at $108.27 on 5 August, 19.8% below it. Peak to trough is (46.3%) in eight weeks.

Realised volatility over those 57 daily returns annualises to 93.5%. The standard error on that estimate is 8.8 percentage points, so the 95% interval runs from 76.2% to 110.8%, a range wide enough to contain almost any view of the stock. Beta needs a market series of comparable length and inherits the same problem. Moving averages, relative strength and every trend measure need a sample this one does not have. The chart shows how the first quarter of trading went, and nothing later in this chapter rests on it.

One price level does carry mechanical weight, and it is in section 9.4.

9.3 The raise, and what it has funded

The offering was entirely primary. No selling shareholder appears in the prospectus, so all $85,675m of net proceeds went to the balance sheet. The underwriters exercised the option to purchase additional shares in full, taking the deal from the 555,555,555 shares on the prospectus cover page, which is the figure the price block quotes, to the 638,888,888 shares settled. The two are the same offering at two stages.

Table 9.1 The June 2026 financings

The Class A voting share of 11.5% is the figure at the offering; on the 28 July 2026 share counts in Table 6.1 it is 12.3%, the difference being Class B conversions in the intervening six weeks.

The prospectus named four uses for the equity proceeds, in this order: expansion of AI compute infrastructure, enhancements to launch infrastructure and launch vehicles, increases in the scale and capacity of the satellite constellations, and general corporate purposes. The 10-Q reports capital additions by segment, so the four uses can be tested against the first quarter of spending.

Table 9.2 Capital additions by segment

Source: 10-Q 0001628280-26-052535, segment note.

AI compute was named first and is taking 86 cents in every dollar of capital deployed in the June quarter, against 48 cents a year earlier. Launch infrastructure and the constellation, the second and third stated uses, together take under 14 cents.

Half year capital expenditure of $28,476m mostly predates the June closings. Of the $110.7bn raised, $40.8bn retired the $20bn bridge loan and other borrowings, $6.4bn went into marketable securities, and $94.4bn of cash, cash equivalents and restricted cash was still on the balance sheet at 30 June. Cash and marketable securities of $100,009m against $39,364m of debt and finance leases leaves the company in the $60,645m net cash position used throughout this report, with a further $5,000m undrawn under the revolver.

For a shareholder, the notes do three things. They fix $1,464m of annual interest cost at a weighted average coupon of 5.855%, a small charge against a company whose interest paid in the half year was $1,667m. They push the first material principal repayment out to July 2031, so the three year projection in section 9.5 carries no refinancing deadline. And they add a covenant capping additional liens at 7.5% of consolidated total assets, which constrains how the next round of capital can be raised if the AI data centre build continues at $16bn a quarter. Financing risk in the scenarios is therefore a risk about the terms of the next raise.

9.4 The lockup as a dated supply timetable

Chapter 7 sets out the release ladder in full. Two points belong here, because both bear on price.

First, the one conditional tranche of the ladder has failed its test. The prospectus releases an additional 455.8 million shares on the second trading day after the first earnings release only if the close was at least 30% above $135.00, that is $175.50, on five of the ten trading days ending on and including that release date. The company furnished its first results on 4 August 2026 (8-K 0001628280-26-052515), which fixes the window. Across those ten sessions the highest close was $125.33, on 4 August itself, so the count was zero of the five required. The outcome is established from the market record rather than from a filing, because no filing reports it: no SEC filing carries daily closing prices and none since reports a release. Those 455.8 million shares stayed inside the 180 day pool and now fall due on 8 December 2026 instead, taking that date's release from 328.4 million shares to 797.6 million, as chapter 7 sets out.

Second, the supply is large against daily trading volume rather than against the total share count. Average volume over the last 20 sessions is 69.3 million shares a day. The 1,230.5 million shares released to date are about eighteen days of that volume. The seven steps between 9 September and 8 December release a further 3,461 million shares, about 50 days of volume, and they land on dates that were printed in a document filed in June. One of the seven is results linked and moves with the reporting calendar; the other six are calendar dates that cannot move.

9.5 Three year scenarios

The model anchors on the June 2026 quarter annualised, because that quarter is the only one reported as a public company and it carries a very different segment mix from the half year. It then projects three years to 2029. Connectivity revenue is built from subscribers and ARPU plus an enterprise and government line grown at a stated rate. Space revenue is built from customer launches at $58m each, the half year rate of $978m of launch services revenue over 17 customer launches, plus development revenue. AI revenue is stated directly. Segment operating margins ramp linearly from the margins the June quarter reported: the Connectivity segment 38.6%, the Space segment (56.3%), the AI segment (49.1%). Depreciation follows the capital base at 19.0% of the average net balance, the rate the June quarter implies. Every input is in Table 9.3, so any case can be rebuilt from this chapter alone.

Table 9.3 Core assumptions, 2029 unless stated

Starting points, all filed: 8.9m subscribers and $81 ARPU at 31 December 2025, 12.0m and $66 at 30 June 2026; 170 launches in 2025 and 78 in the first half of 2026, of which 17 carried a customer primary payload; nameplate compute draw of 1.4 GW at 30 June 2026 against 0.4 GW a year earlier; revenue of $18,674m in 2025 (424B4 0001628280-26-042639; 10-Q 0001628280-26-052535).

Table 9.4 Scenario outputs, 2029

The base case is an estimate under the assumptions in Table 9.3, and three things break it. Subscriber growth of 101% year on year in the June quarter slowing faster than the 22% ARPU decline slows, which would turn the Connectivity segment from a fast growing business into a mature one. The single customer at 18.3% of June quarter revenue across all three segments, 20.9% of 2025 revenue and 25.2% of 2023 revenue, not renewing on comparable terms (424B4 0001628280-26-042639 note 3; 10-Q 0001628280-26-052535 note 3). And AI revenue failing to follow nameplate compute: the segment more than tripled revenue year on year in the June quarter, from $737m, while still losing $1,257m on $2,561m of revenue, and the base case asks it to reach $28,000m and a positive margin by 2029.

The scenarios use no EBITDA anchor. With capital expenditure running above revenue and depreciation of $30,599m against revenue of $69,361m in the 2029 base case, the calculation produces EBITDA margins of 60% to 85%. The terminal metric is enterprise value over revenue throughout.

9.6 What has to be true for the best case

Four conditions carry it, and each is testable against a number the company already reports.

Starship reaching customer cadence. The best case needs 110 customer launches in 2029 against 17 in the first half of 2026. Falcon flew 77 times in that half year but only 17 carried a customer primary payload; the other 60 were internal. Starship flew once. Space segment revenue tracks customer launches, so cadence alone does not convert into revenue until third party payloads displace internal ones.

AI revenue catching up with installed compute. Nameplate compute draw went from 0.4 GW to 1.4 GW in a year while AI revenue went from $737m to $2,561m a quarter. The best case needs $52,000m by 2029 at a 28% operating margin. The capital is already committed and depreciating, so the swing factor is monetisation per gigawatt.

ARPU holding while subscribers compound. The best case is the only one in which ARPU stabilises, at $62 against $66 in the June quarter and $81 in 2025. Management has said it prioritises subscriber growth and cost reduction over ARPU and expects ARPU to keep declining as the base shifts outside North America (424B4 0001628280-26-042639, key business metrics). The best case therefore requires a mix or pricing outcome management has not guided to.

Capital intensity falling to 0.42 times revenue by 2029. It was 1.11 in 2025 and 2.28 in the first half of 2026. This is what converts the story into free cash flow, and it happens only if AI capital spending plateaus while AI revenue keeps compounding.

9.7 Sensitivity

Table 9.5 One variable at a time against the base case of $107.07 per share

The terminal multiple moves the answer nearly twice as far as the largest operating driver, and more than four times as far as the six smallest drivers in the table combined. Capital intensity and the segment margins barely register, because a revenue multiple only picks them up through year end net debt. Under an earnings or cash flow anchor they would rank far higher, so the multiple is the assumption this section examines.

The terminal multiple is a judgement rather than a filed figure. SpaceX has three audited financial years in a registration statement and one quarter as a public company. It has no annual report, no proxy statement and no multiyear track record of guidance against delivery. There is no clean listed comparator, as chapter 4 sets out. A terminal multiple applied to 2029 revenue on that evidence base is a judgement about a company that will have filed three annual reports by then, none of which exists today.

Across the two way grid of multiple against revenue the implied value runs from $20 to $301 per share. Holding 2029 revenue at the base case of $69,361m and moving only the multiple from 8.0x to 38.0x moves the answer from $40 to $184, a factor of 4.6. Holding the multiple at the base 22.0x and moving 2029 revenue across its full range, from the worst case $34,160m to the best case $113,820m, moves it from $54 to $175, a factor of 3.3. The multiple wins even against a revenue range that spans a 3.3 times difference in the size of the business.

The $1,973.84bn market value in section 9.1, less the $60.6bn net cash position at 30 June 2026, is an enterprise value of $1,913bn. Divided by the base case 2029 revenue of $69,361m, that is a terminal multiple of 27.6 times. The market is therefore paying between the base case multiple of 22.0 times and the best case multiple of 38.0 times, on the base case's own revenue path. Whether that is expensive turns on which of Table 9.3's revenue columns comes true and on what a company with three audited years should trade at. Neither is knowable from what has been filed.

The ordering in Table 9.5 carries more of this chapter's content than any single figure in Table 9.4. The scenarios are a range under stated assumptions. They are not a target and they are not advice.

Sources and credits

Every SEC filing used in this report was retrieved through the SEC-API.io MCP server. The registrant is Space Exploration Technologies Corp., CIK 1181412, ticker SPCX.

Market prices carry no named data provider. Prices are last sale prices from public market data, struck at the close of trading on Thursday 3 September 2026, and they sit outside the SEC-API.io credit. Share counts, and the cover pages they come from, are SEC filings and sit inside it.

Table A.1 The filings this report is built on

This is an independent analysis derived from SEC filings. It is not a publication of the Securities and Exchange Commission, and the SEC has neither reviewed it nor endorsed it. The abbreviation SEC names the source of the filings throughout this report, never its author.

Disclaimer

This report is not financial advice. It is an independent analysis of public SEC filings, prepared for information only. Nothing in it is a recommendation, an offer or a solicitation to buy or sell any security, and nothing in it is investment, legal, tax or accounting advice. It takes no account of the objectives, financial situation or needs of any reader.

Three features of this particular company make that warning more than a formality. SPCX listed on 12 June 2026, twelve weeks before the date of this report, and its entire audited record is three years of statements inside one prospectus, the 424B4 of 12 June 2026. The company has filed no annual report on Form 10-K and no proxy statement, and its only periodic report is a single quarterly report on Form 10-Q, for the quarter ended 30 June 2026. That leaves no history of guidance against delivery to test management by. The scenario chapter spans $19 to over $306 per share on stated assumptions, and the sensitivity table shows the terminal multiple, a judgement rather than a filed figure, moving the answer further than any operating driver; those are illustrations of what the assumptions imply, not forecasts, targets or price objectives. And the lockup releases several billion further shares on dates already fixed, which is a supply fact and not a view on price.

Figures are as filed, and filings are restated, amended and superseded. The market prices used here are as at the close on 3 September 2026 and are stale the moment the market next opens. Past performance, and twelve weeks of it least of all, does not indicate future results. Anyone acting on any part of this report should verify the underlying filings and take advice from a qualified professional. The author holds no position in any security named here and receives no compensation from any company named here.