SpaceX

Eleven weeks in, history's biggest IPO has already lived a full boom-bust-recovery cycle — up 60%, down 20%, settling at +5% — while nearly everything the S-1 actually promised remains untested.

SpaceX priced at $135 on June 11, 2026, raising roughly $75 billion — the largest IPO ever — with an unprecedented ~30% of shares reserved for retail investors whose orders reportedly topped $100 billion. The scarcity mechanics the S-1 itself warned about took over immediately: a 22% first-day pop, then a vertical run to a $216.56 close (intraday ~$225) on June 16 as SpaceX exercised its pre-IPO option to buy Cursor for $60 billion in stock and Nasdaq fast-tracked the shares toward the Nasdaq-100. Market cap briefly crossed $3 trillion, making it the fourth-most-valuable company in the world within three trading sessions.

Then the risk factors started printing. Institutions publicly balked at Musk's super-voting control, the controlled-company exemptions, and the arbitration/Texas-forum stack — SPCX became the most shorted large public company in America, with short sellers sitting on ~$15.5 billion of paper profits by late July. Operations supplied the catalysts: a July 16 Starship launch abort at T-0 erased roughly $100 billion of market value in a day, a Falcon 9 suffered a rare post-ignition pad abort on July 20, and when Flight 13 finally flew on July 24 it deployed the first 20 Starlink V3 satellites and made the softest-ever ship splashdown — but the booster was destroyed during its landing burn. With ~$116 billion of insider shares due to unlock August 6, the stock bottomed at $108.07 on July 31, 52% off the peak and 20% below the IPO price, ignoring a wall of bullish sell-side initiations along the way.

August inverted the story. The first-ever earnings report (August 4) beat on nearly everything — Q2 revenue of $7.8 billion, up 92%; Starlink at 12 million subscribers, double a year earlier; a net loss of just $541 million — but the stock initially fell on an $18.4 billion quarterly capex shock, roughly $16 billion of it AI compute. Two days later the feared unlock arrived and the stock rose 6%: the bigger float let index funds and institutions finally build positions ahead of the September Nasdaq-100 rebalance. Morgan Stanley's note that the market was 'significantly undervaluing' the AI business reframed the capex as an asset, SPCX reclaimed its IPO price on August 10, absorbed a second $46 billion unlock without incident on August 20, and closed August 28 at $141.50, up 4.8% from the IPO.

What has not happened yet is almost everything the S-1 actually promised. Starship has flown once since the IPO with mixed results; the first orbital flight and tower catch slipped to September. The EchoStar spectrum deal doesn't close until late 2027, orbital AI compute is a 2028 aspiration, and Starlink ARPU fell from $85 to $66 as growth skewed international. Meanwhile the government-concentration risk inverted into a tailwind — roughly $7 billion of 2026 Pentagon awards, including ~91% of Golden Dome money awarded so far. Eleven weeks is long enough to validate the S-1's market-structure warnings in full, and far too short to grade its physics.

SpaceX asks investors to buy the world's only vertically integrated space-connectivity-AI conglomerate: a launch monopoly (over 80% of world mass-to-orbit, 99%+ Falcon success rate) that funds Starlink's fast-compounding subscriber business (10.3 million subscribers, up ~105% year over year, $11.4 billion of 2025 Connectivity revenue), which in turn bankrolls the newly merged xAI segment — Grok, the X platform, and gigawatt-scale COLOSSUS data centers. The 2025 consolidated picture is $18.7 billion of revenue and $6.6 billion of Adjusted EBITDA against a $4.9 billion net loss, with the losses concentrated in AI investment. The forward story is Starship: full, rapid reusability unlocks V3 Starlink satellites, direct-to-cell service on EchoStar spectrum, orbital AI compute powered by the Sun, a lunar economy, and ultimately Mars. Class A buyers get one vote per share; Mr. Musk and Class B holders keep ten votes per share and a charter-guaranteed majority of the board.

The bull case as the S-1 argues it: SpaceX has 'cracked the code' twice — first on cheap, reusable, high-cadence launch, then (via xAI) on building coherent gigawatt-scale AI compute faster and cheaper than anyone. Starship completes flight testing and begins delivering payloads to orbit in 2026, enabling V3 satellites with one Tbps of downlink capacity each and a twenty-fold jump in deployed Starlink capacity per launch. The EchoStar spectrum purchase plus V2 Mobile satellites turn every unmodified phone into a Starlink device, collapsing mobile dead zones worldwide. Starlink cash flow, real-time X data, Grok frontier models, Terafab chip manufacturing with Tesla and Intel, and Cursor's coding franchise compound into a self-reinforcing 'innovation engine.' By 2028, orbital AI compute satellites begin harvesting effectively limitless solar power, and the same launch infrastructure opens trillion-dollar markets that do not yet exist: the lunar economy, asteroid mining, and cities on Mars.

With hindsight: The near-term, checkable half of the bull case is largely on track: revenue grew 92% year over year in the first post-IPO quarter, Starlink subscribers doubled to 12 million, Starship deployed the first 20 V3 satellites on Flight 13 (suborbital), the AI segment grew 247%, and compute demand was strong enough that Musk declared SpaceX 'exclusive to Nvidia' and pulled the internal $1 trillion revenue target forward to 2030. But the load-bearing claims are unproven or slipping: Starship has not yet delivered a payload to orbit (Flight 14, with the first tower catch, slipped from August to September), full reusability remains undemonstrated after Booster 20 was destroyed, Gen2 direct-to-cell awaits the 2027 EchoStar close, orbital AI compute is still a 2028 promise, and Starlink ARPU fell from $85 to $66. Eleven weeks in, the flywheel is spinning but none of the step-function claims — the ones the valuation rests on — can yet be graded.

• Starship completes flight testing and begins delivering payloads to orbit in 2026, with expected performance, reusability, and cost efficiencies.

  “the development and deployment of Starship in accordance with our anticipated schedule (including commencement of payload delivery to orbit in 2026) and launch cadence and our ability to achieve expected performance, reusability, and cost efficiencies” source ↗

• Orbital AI compute satellites could begin deploying as early as 2028.

  “the deployment of our next-generation Starlink satellites, satellite-to-mobile connectivity, and orbital AI compute infrastructure (including potential deployment of our orbital AI compute satellites as early as 2028)” source ↗

• SpaceX has historically outperformed competitors in its Space and Connectivity segments — the track record underwriting the whole flywheel.

  “While we have historically outperformed certain competitors in aspects of our business, such as our Space and Connectivity segments, there can be no assurance that we will maintain this position.” source ↗

• Starlink Mobile Gen2 will deliver 5G-like connectivity to existing, unmodified phones in the United States using V2 Mobile satellites and the EchoStar spectrum.

  “In the United States, we expect to be able to provide 5G-like connectivity to a meaningful portion of existing unmodified devices through our Starlink Mobile Gen2 service utilizing our V2 Mobile satellites” source ↗

• Excess AI compute is already being monetized externally — including a multi-year cloud deal signed with Anthropic weeks before the filing.

  “On May 3, 2026, the Company entered into a cloud services agreement with Anthropic PBC, an AI research and development public benefit corporation, with respect to access to compute capacity.” source ↗

1. Everything hinges on Starship

   partly came true

   growth · structural

   The entire growth strategy — V3 Starlink satellites, Gen2 direct-to-cell, orbital AI compute, lunar and Mars ambitions — requires Starship to achieve full reusability, high cadence, and orbital payload delivery on schedule. Falcon 9 and Falcon Heavy physically cannot deploy the next-generation satellites, and the filing concedes AI compute satellites at scale need full Starship reusability to be economically compelling.

   “If we are unable to successfully complete the development, testing, and deployment of Starship at scale in accordance with our anticipated schedule, or at all, or if we are unable to achieve sufficient launch cadence, reusability, and capability, our ability to execute our growth strategy” source ↗

   What happened: The risk showed its teeth without fully materializing: a July 16 T-0 launch abort erased ~$100 billion of market value in one day, and Flight 13 (July 24) deployed the first 20 Starlink V3 satellites but lost its booster during the landing burn. Flight 14 — the first orbital flight and first tower-catch attempt — slipped from August to NET September 2026, so 'payload delivery to orbit in 2026' remains achievable but unproven.

2. Dependence on Elon Musk, a part-time CEO

   didn't happen

   key person · structural

   SpaceX is explicitly dependent on Musk's leadership and culture-setting, carries no key-person life insurance on him, and admits he does not devote his full time to the company — he simultaneously runs Tesla and is involved in Neuralink and The Boring Company. His statements and outside ventures can move the stock in either direction.

   “Further, although Mr. Musk devotes significant time to our businesses and is highly active in our management, he does not devote his full time and attention to our businesses and devotes time and attention to other significant roles” source ↗

   What happened: No loss or reduction of Musk's involvement occurred — he led the first earnings call, set targets, and taunted short sellers on X. The Musk factor that did hurt the stock in this window was his voting control (a governance issue), not his absence or divided attention.

3. Dual-class structure hands Musk and Class B holders control of the board

   came true

   governance · structural

   Class B stock carries ten votes per share, and Class B holders voting as a separate class elect 51% of the board and can remove those directors. Musk — irremovable as CEO and Chairman except by Class B vote — controls a majority of Class B. SpaceX is a 'controlled company' that intends to use exemptions from independent-committee requirements. Class A buyers are structurally voiceless.

   “Under our charter, holders of our Class B common stock, voting separately as a class, will be entitled to elect 51% of the total number of authorized directors constituting our board (rounded up to the nearest whole number) and will have the ability to remove those directors” source ↗

   What happened: The disclosed 'adverse impact on the trading price' arrived within weeks: institutional investors publicly cited Musk's >80% voting power, the controlled-company exemptions, and irremovable CEO/Chairman roles as reasons to stay away, and SPCX became the most shorted large public company in America during the July slide to $108.

4. Musk and his affiliates are free to compete and take corporate opportunities

   unclear

   governance · structural

   The charter affirmatively renounces corporate opportunities presented to Musk and certain directors, and permits Musk-affiliated entities to compete with SpaceX outright. The company is already deep in related-party dealings: Terafab with Tesla (terms not finalized), Tesla Megapack and Cybertruck purchases, and roughly $9 billion of sale-leaseback debt with Valor Equity Partners, whose founder sits on the board.

   “Under our charter, Mr. Musk and his affiliates are not restricted from owning assets or engaging in businesses that compete directly or indirectly with us and will not have any duty to refrain from engaging, directly or indirectly, in the same or similar business activities or lines of business as us” source ↗

   What happened: No new conflict-of-interest incident surfaced in the window; the Cursor option was exercised as pre-arranged and the Tesla/Valor arrangements continued as disclosed. Whether the renounced-opportunities charter costs Class A holders anything will take years to observe.

5. AI segment capital intensity with no assured return

   came true

   profitability · structural

   The AI segment lost $6.4 billion from operations in 2025 while consuming $12.7 billion of capex ($7.7 billion in Q1 2026 alone), and the filing warns of a multi-year investment horizon before sustained positive Segment Adjusted EBITDA — while conceding the commercial value of frontier models is largely unproven and scaling laws may not hold.

   “Our AI segment has incurred significant operating losses since inception, and we may not achieve profitability in this segment, or, if achieved, sustain it, and there can be no assurance that the returns on our AI investments will be adequate to justify the capital deployed.” source ↗

   What happened: Q2 2026 capex hit $18.4 billion — roughly $16 billion of it AI compute, versus ~$13.2 billion expected — and the stock fell 5-8% after hours on the disclosure despite a revenue beat. The market later reframed the spend after Morgan Stanley's 'undervalued AI business' note, but the capital intensity the filing warned about is fully in evidence; returns remain unproven.

6. Orbital AI compute has never been attempted by anyone

   unclear

   tech & security · structural

   The plan to put AI data centers in orbit — potentially up to one million satellites — is untested by SpaceX or anyone else. Hardware cannot be repaired once deployed, satellites live shorter lives than the computers they host, and the filing admits the whole concept could be mooted by cheap terrestrial energy such as nuclear breakthroughs.

   “In particular, we have not, and no one else has, previously operated or attempted to operate orbital AI compute, and the conditions of space on such AI infrastructure have not been tested.” source ↗

   What happened: No orbital AI compute hardware has been deployed; the S-1's own timeline puts first deployments at 2028 at the earliest. Nothing in the first eleven weeks tests this risk either way.

7. A history of multibillion-dollar net losses

   partly came true

   profitability · serious

   Despite $18.7 billion of 2025 revenue and positive consolidated Adjusted EBITDA, SpaceX has never been reliably profitable: a $4.9 billion net loss in 2025, $4.3 billion lost in Q1 2026 alone, and a $41.3 billion accumulated deficit, with capital expenditures still accelerating.

   “We incurred net losses of $(4,937) million and $(4,628) million for the years ended December 31, 2025 and 2023, respectively, and a net loss of $(4,276) million for the three months ended March 31, 2026.” source ↗

   What happened: SpaceX remained unprofitable in its first reported quarter as a public company, but the Q2 net loss of $541 million was dramatically smaller than the $4.3 billion Q1 loss and beat expectations ($0.09 loss per share vs. $0.26 expected), with adjusted EBITDA of $3.5 billion, up 191%.

8. xAI integration is incomplete months after the merger

   unclear

   operations · serious

   xAI was acquired in February 2026 — barely three months before the S-1 — and was itself an early-stage company. SpaceX admits the management structure, controls, and procedures needed to run the combined company are still being put in place, and that Grok usage metrics can swing significantly with competitor releases.

   “we have had to take significant steps to integrate xAI’s operations into our broader corporate structure as part of our AI segment, including putting in place the management team and organizational structure needed to execute at the scale and pace our strategy demands, as well as controls and procedures appropriate for a larger organization like ours” source ↗

   What happened: No integration failures surfaced publicly; AI segment revenue grew 247% year over year in Q2. But the $60 billion Cursor acquisition announced June 16 (expected to close in Q3) adds a second major integration on top of the unfinished xAI one, and controls/procedures work disclosed in the S-1 has had no public test yet.

9. FAA licensing and mishap investigations gate launch cadence

   partly came true

   regulation · serious

   Launch and reentry require FAA licenses; current rules do not even permit return-to-launch-site Starship reentries without a waiver. Anomalies trigger investigations and groundings — as has happened before — and FAA resource strain could slow the very cadence increases the business plan requires.

   “Following an anomaly, mishap, or failure, the FAA or other authorities may require investigations, impose corrective actions, or restrict or delay our ability to conduct launch operations. We have in the past been, and may in the future become, subject to such actions, impacting our ability to increase launch cadence.” source ↗

   What happened: Flight 14's first-ever tower catch of the Starship upper stage was gated on FAA approval of an expanded landing-zone safety profile, and the flight slipped from 'as soon as August' to NET September 2026. No license denial or grounding occurred, but regulatory pacing visibly shaped the schedule exactly as disclosed.

10. Gen2 direct-to-cell depends on the EchoStar spectrum deal and global approvals

    unclear

    regulation · serious

    The $19.6 billion EchoStar spectrum purchase does not close until November 2027, country-by-country authorizations are pending nearly everywhere, no commercially available handset yet supports the relevant 5G NTN bands, and some countries are openly considering ignoring the ITU priority rights SpaceX is buying.

    “we must secure the global right to use the spectrum acquired from EchoStar from a number of international telecommunications regulators in order to make our V2 satellite-to-mobile services usable worldwide, and there can be no assurance that such authorizations will be granted on acceptable terms, or at all.” source ↗

    What happened: The spectrum transaction is not due to close until November 2027, so the core risk is untested. Commercial momentum was positive in the window — direct-to-cell launched in Italy and Starlink signed its largest-ever D2C deal with Veon (~150M+ potential customers) — while Amazon's Leo filed with the FCC for a rival 5,105-satellite direct-to-device constellation.

11. Grok content-safety investigations and litigation

    unclear

    legal · serious

    Grok's deliberately 'less constrained' modes have already produced regulator inquiries on multiple continents (Irish DPC, FTC), putative class actions, and a Baltimore consumer-protection suit over nonconsensual explicit imagery — filed against SpaceX itself as well as the xAI entities — plus copyright litigation over model training.

    “we are subject to investigations and inquiries from regulators and law enforcement authorities in the United States and internationally concerning allegations that our AI products were used to create nonconsensual explicit images or content representing children in sexualized contexts, and similar matters.” source ↗

    What happened: The pre-IPO putative class actions, the Baltimore suit, and the Irish DPC and FTC inquiries disclosed in the filing remain pending; no new enforcement action, ruling, or fresh scandal was reported in the June-August window.

12. U.S. government revenue concentration and shifting political priorities

    didn't happen

    other · serious

    Roughly a fifth of revenue comes from U.S. federal agencies under mostly fixed-price contracts the government can terminate at its convenience, with funding hostage to appropriations and a polarized political environment. SpaceX also warns it may prioritize its own payloads over government customers to hit orbital-compute goals — inviting litigation and regulator friction.

    “In 2025, approximately one-fifth of our revenue was attributable to agencies within the U.S. federal government.” source ↗

    What happened: The risk ran in reverse: SpaceX won roughly $7 billion of 2026 Pentagon awards, including ~$6.45 billion of Golden Dome contracts in late May and all 18 Space Force SBST Falcon 9 missions ($1.6 billion) in late July — reportedly ~91% of Golden Dome money awarded so far. Concentration deepened, but as a tailwind, not a harm.

13. Exposure to hostile foreign governments and asset seizure

    didn't happen

    market · serious

    Starlink's global footprint exposes it to capricious regimes: Brazil's Supreme Court froze Starlink's assets in 2024 over conduct by X — then not even owned by SpaceX — and foreign governments have publicly discussed using anti-satellite weapons against the constellation. The filing warns similar Musk-affiliation spillovers can recur anywhere.

    “in August 2024, Starlink received an order from Brazil’s Supreme Court that froze Starlink’s Brazilian financial assets and prevented Starlink from conducting financial transactions in Brazil” source ↗

    What happened: No asset seizure, expropriation, or anti-satellite incident was reported in the window; Starlink expanded commercially (Italy D2C launch, Veon deal) rather than losing markets.

14. $29 billion of debt on a loss-making company

    didn't happen

    financing · serious

    Total principal indebtedness stood at $29.1 billion at March 31, 2026, much of it at variable rates and swollen by AI infrastructure sale-leasebacks; covenant restrictions and the possibility of further equity raises to fund capex mean dilution and refinancing risk sit alongside the operating risks.

    “As of March 31, 2026, we had total principal indebtedness outstanding of $29,132 million” source ↗

    What happened: No financing distress materialized — the ~$75 billion IPO raise transformed the liquidity picture, and no covenant issues, downgrades, or new emergency financings were reported despite $18.4 billion of quarterly capex.

15. Shareholder rights stripped: Texas forum, arbitration, no class actions, no juries

    partly came true

    governance · serious

    The bylaws route all shareholder disputes — including federal securities claims — to the Texas Business Court, fall back to mandatory ICC arbitration, prohibit class or collective actions, impose a 3% ownership threshold for derivative suits and a 67% solicitation requirement for proposals, and deem every buyer to have waived jury trial.

    “our bylaws will provide that any person or entity purchasing or otherwise acquiring or holding any interest in shares of stock of the Company shall be deemed to have irrevocably and unconditionally waived any right it may have to a trial by jury in any Internal Dispute.” source ↗

    What happened: The provisions have not yet been tested in litigation, but they imposed a real cost sooner than expected: mandatory arbitration and Texas-law barriers to derivative suits were among the specific grievances institutional investors cited during the July governance revolt that helped drive the stock 20% below its IPO price.

16. A record retail allocation that invites volatility

    came true

    market · serious

    SpaceX explicitly warned that an unusually large share of the offering was earmarked for retail investors and that high retail interest could amplify swings in the stock — a self-diagnosis of the trading dynamics its own offering structure created.

    “a number of shares of our Class A common stock are expected to be allocated to retail investors in this offering. Additionally, high retail investor interest in our Class A common stock may occur following this offering, which may lead to increased volatility of the trading price.” source ↗

    What happened: Textbook. The ~30% retail allocation and >$100 billion of retail demand on a small float produced a +60% spike in three sessions, a 52% peak-to-trough collapse to $108 by July 31, and an August round trip back through the IPO price — at which point data showed retail investors ending their buying spree. The S-1's self-diagnosis was precise.

How the S-1 reads

This S-1 reads like two documents fused together: a swaggering operating history ('cracked the code,' 'The Algorithm,' 80% of world mass-to-orbit) and an unusually blunt risk section that admits key markets do not exist, orbital AI has never been attempted by anyone, and management cannot fully assess its own risks because the ventures are unprecedented. The candor is real — named past failures, the Brazil seizure, Grok's 'Unhinged' mode, unfinished internal controls — but it coexists with the most shareholder-hostile governance stack of any mega-cap IPO: ten-vote Class B shares electing 51% of the board, renounced corporate opportunities, controlled-company exemptions, and a Texas forum/arbitration/jury-waiver regime built on 2025 Texas statutes that the filing itself expects to be challenged. The document's deepest tell is structural: the rocket company is the smallest segment by revenue, and the offering substantially funds an AI capex race that was bolted on ninety days before filing.

• “Our plans to deploy large-scale orbital infrastructure, including orbital AI compute systems, will require the operation of very large satellite constellations, potentially numbering up to one million satellites.” source ↗

  The scale of the orbital AI ambition, stated flatly in a risk factor: a constellation two orders of magnitude beyond today's Starlink.

• “Certain of our AI products, including Grok, offer features or modes designed to generate more candid, direct, or less reserved or irreverent outputs, such as “Spicy” Imagine Mode and “Unhinged” Voice Mode.” source ↗

  An S-1 first: the issuer formally disclosing its chatbot's 'Spicy' and 'Unhinged' modes as risk factors.

• “Mr. Musk currently serves as Technoking and Chief Executive Officer of Tesla and is involved in other emerging technology ventures, including Neuralink and The Boring Company.” source ↗

  The key-person risk factor states, in SEC-filing prose, that the CEO of history's largest IPO holds the title 'Technoking' at another public company.

• “Others, including in-orbit manufacturing, passenger transport to the Moon, an established human presence or gateway hub on the Moon, passenger and cargo transport to Mars, energy production on the Moon or Mars, manufacturing capabilities on the Moon or Mars, and asteroid mining do not exist today.” source ↗

  Rare candor about the addressable market: much of the long-term story is markets that literally do not exist.

• “Cursor is entitled to a $1.5 billion termination fee under the option agreement and an $8.5 billion deferred services fee under the compute agreement” source ↗

  The Cursor option's break-fee structure — a $10 billion cost of changing its mind, disclosed as a subsequent event.

What this one teaches

• An S-1 can correctly predict its own trading: SpaceX's retail-allocation volatility warning was the single most accurate sentence in the filing, playing out as a +60%/-52% round trip inside eight weeks.
• Governance terms are not abstractions — dual-class control, controlled-company exemptions, and arbitration bylaws were the named reasons institutions shorted rather than bought, imposing a measurable discount long before any court tests the provisions.
• In a conglomerate S-1, follow segment capex, not the cover story: the 'rocket company' spent $16 billion in one quarter on AI compute, and the stock traded on that number, not on launches.
• Feared mechanical events beat feared narratives: the $116 billion lockup expiration everyone dreaded produced a +6% day, while the risks that actually moved the stock — a T-0 abort, a capex print — were the ones no calendar flagged.
• When a filing says its ventures are 'unprecedented' and its markets 'do not exist today,' believe it in both directions: eleven weeks was enough to grade the market-structure disclosures and almost none of the physics.

The paper trail

1. 2026-05-20 S-1 filing index ↗ document ↗
2. 2026-06-01 S-1/A filing index ↗ document ↗
3. 2026-06-03 S-1/A filing index ↗ document ↗
4. 2026-06-12 424B4 filing index ↗ document ↗

Filed as SPACE EXPLORATION TECHNOLOGIES CORP. All documents are public domain, served by SEC EDGAR.