SpaceX IPO
SpaceX just filed its S-1. The document is exactly what you'd expect from a 24-year-old company that has become the most important private business in the world: enormous, ambitious, frequently surreal, and structured in a way designed to ensure that one person stays in complete control of the whole thing forever.
Let me be clear: I read the entire filing. This is a deep-dive into the S-1 of Space Exploration Technologies Corp., and what it reveals is a company that is simultaneously the most valuable private enterprise in history and a thing of profound financial strangeness.
Here's everything the S-1 tells us — and everything it quietly hides.
THE TOP-LINE NUMBERS
SpaceX generated $18.674 billion in revenue in 2025, up from $14.015 billion in 2024 and $10.387 billion in 2023. That's a three-year compound annual growth rate of roughly 34%. Impressive. Unquestionably impressive.
But the picture gets stranger when you look at profitability. The company reported a loss from operations of $2.589 billion in 2025, swinging from a $466 million operating profit in 2024. Net loss for 2025 was $4.937 billion. The Q1 2026 numbers are even uglier on the surface, with a $1.943 billion operating loss and a $4.276 billion net loss on $4.694 billion in revenue.
The reason for the swing is simple: xAI. SpaceX completed its acquisition of xAI in February 2026, and the financials have been retrospectively recast to combine the two companies for all periods presented. That means the 2023 and 2024 numbers you'll see in financial press coverage are not what SpaceX previously disclosed — they've been redone to include xAI's losses, which makes the historical profitability picture look worse than the actual rocket-and-satellite business is.
Apples-to-apples, the core SpaceX business is doing significantly better than these headline numbers suggest. The losses are an accounting artifact of absorption, not a sign of deterioration.
THREE BUSINESSES UNDER ONE ROOF (AND THEY ARE NOTHING ALIKE)
SpaceX now reports in three segments, and the differences between them are enormous:
| Segment | 2025 Revenue | 2025 EBITDA | Margin |
|---|---|---|---|
| Connectivity (Starlink) | $11.39B | $7.17B | 63% |
| Space (Launch) | $4.09B | $0.65B | 16% |
| AI (xAI + X) | $3.20B | -$1.24B | Negative |
Connectivity is the actual cash machine. Starlink generated $11.4 billion in revenue in 2025, up 49.8% year over year, and threw off $7.17 billion in Segment Adjusted EBITDA at a 63% margin. Connectivity EBITDA grew 86.2% year over year. 10.3 million subscribers. This is a broadband business disguised as a satellite company, and it's worth 80% of SpaceX's entire revenue.
The Space segment is profitable but tiny. Rocket launches to NASA, the Department of Defense (which the filing refers to as the "Department of War"), and commercial customers. $4.1 billion in revenue. It's the foundational business — the reason any of this is possible — but it's no longer the growth story.
The AI segment is a black hole. xAI plus X (formerly Twitter) generated $3.2 billion in revenue and lost $6.36 billion at the operating line. This is the segment that dragged the entire consolidated picture into the red. Starlink's success is effectively subsidizing xAI's expenditures at a scale that is hard to comprehend.
THE ANTHROPIC PLOT TWIST (THIS IS THE ONE THAT GOT ME)
Buried in the "Recent Developments" section of the S-1 is this disclosure:
*"In May 2026, we entered into Cloud Services Agreements with Anthropic PBC, an AI research and development public benefit corporation, with respect to access to compute capacity across COLOSSUS and COLOSSUS II. Pursuant to these agreements, the customer has agreed to pay us $1.25 billion per month through May 2029."*
Let me restate that. Anthropic — which makes Claude and is xAI's direct competitor in frontier AI models — has signed a contract to pay SpaceX $1.25 billion per month for compute capacity. That contract runs from June 2026 through May 2029, which works out to approximately $45 billion in total contract value over the 36-month term.
There are a few ways to read this. The simplest is that xAI has built out so much compute capacity at its COLOSSUS and COLOSSUS II data centers in Memphis and Mississippi that it can't use it all internally, and renting the spare capacity to Anthropic is a way to monetize the overhead. The S-1 calls it a "dual monetization strategy" that provides "multiple pathways to generate returns on invested capital."
The other way to read it is that Anthropic is willing to write checks of this size because GPU and data center capacity is so scarce that even your direct competitor's infrastructure is worth paying for. The agreement can be terminated by either party on 90 days' notice, which is interesting given the size of the dollar figure involved.
Either way, the optics are unusual: SpaceX's IPO prospectus is showing investors a massive AI compute revenue stream that is coming from xAI's largest competitor.
THE CURSOR OPTION (AND THE $10 BILLION BREAKUP FEE)
A second weird disclosure that hasn't gotten much attention: in April 2026, SpaceX entered into a compute agreement plus an option agreement with Anysphere, Inc. — the company behind Cursor, the popular AI-powered code editor.
Under the option agreement, SpaceX has the right (but not the obligation) to acquire Cursor at a predetermined price implied at a $60 billion equity value. If SpaceX decides not to exercise the option, or if Cursor terminates because of a SpaceX material breach, the breakup mechanics kick in:
- $1.5 billion termination fee under the option agreement
- $8.5 billion deferred services fee under the compute agreement
- Total potential breakup payment: $10 billion
These fees are payable in cash, or in Class A common stock if the IPO has not closed when they become payable. So Cursor is essentially saying: *"We'll let you have the option to buy us at $60 billion, but if you walk away, you owe us $10 billion."* That is a remarkable structure for what is technically a software development tools company.
THE TESLA / SPACEX / X / BORING CO. MONEY WEB
The S-1's related-party transactions section (Note 18 in the financial statements) is the section that institutional fiduciaries have been most alarmed by. It discloses the actual dollar flows between SpaceX and the other companies Musk leads. The numbers are bigger than expected.
| Transaction | Amount | Period |
|---|---|---|
| Tesla cash investment in xAI (Series E Preferred) | $2,000M | January 2026 |
| xAI/Tesla goods (Megapack products) | $506M | FY 2025 |
| xAI/Tesla goods (Megapack products) | $191M | FY 2024 |
| SpaceX/Tesla commercial agreements | $144M | FY 2025 |
| SpaceX/Tesla commercial agreements | $4M | FY 2024 |
| SpaceX invoiced Tesla for aircraft use | $2M | FY 2025 |
| SpaceX paid Boring Co. for Bastrop tunnels | $1M | FY 2025 |
| xAI lease to Musk Industries LLC | $2M | FY 2025 |
| X lease payments to Boring Company | $1M | FY 2025 |
The most consequential of these is the first one. In January 2026, Tesla agreed to invest $2 billion in xAI through a purchase of Series E Redeemable Convertible Preferred Stock. The transaction was subject to regulatory approval, which came through after the SpaceX-xAI merger closed. As a result, on March 12, 2026, Tesla received 3.8 million shares of SpaceX Class A common stock (on a pre-split basis) in place of the original xAI preferred stock.
Restated in plain language: Tesla shareholders' money was used to buy SpaceX equity at a specific implied valuation, before SpaceX had public shareholders or an independent compensation committee to evaluate whether the deal terms were fair. The two companies share a CEO. The CEO sits on both boards. The deal happened entirely within Musk's ecosystem of controlled entities.
Pension fund signatories of the May 13 Comptroller letter (CalPERS, NYC Comptroller, and NYS Comptroller jointly) cited this transaction specifically as evidence of the governance concerns they wanted addressed before the IPO.
The xAI/Tesla goods purchases are also worth pausing on. xAI bought $506 million of Megapack battery products from Tesla in 2025, growing from $191 million in 2024. Megapacks are utility-scale lithium-ion batteries used to stabilize power grids and provide backup power to data centers. The xAI/Tesla flow effectively converts Tesla manufacturing capacity into xAI compute infrastructure, with both companies benefiting from the related-party flow but with the customer/supplier relationship priced inside Musk's controlled ecosystem.
The aircraft billing line is the most unusual disclosure. SpaceX owns and operates aircraft that are used by Musk in his capacity as CEO of Tesla, and bills Tesla for that usage. $2 million in 2025.
Two additional disclosures from this section are worth flagging: First, SpaceX paid The Boring Company $1 million in 2025 for tunnel construction in Bastrop, Texas (where SpaceX has its main Starlink manufacturing facility). Second, xAI leases real property from "Musk Industries LLC," an entity owned by Mr. Musk directly. Under this arrangement, xAI paid $500,000 in 2024, $2 million in 2025, and $200,000 in the first two months of 2026. A separate Musk-controlled entity is therefore receiving rent from a subsidiary of the company that Musk also controls.
None of these flows are individually material to the consolidated financial statements, but the pattern of cross-billing among Musk-controlled entities is exactly what the May 13 Comptroller letter flagged as a governance concern.
THE INTANGIBLE ASSET PROBLEM
One quieter consequence of the xAI/X consolidation is that SpaceX now carries a large book of intangible assets that are not aerospace-related. The S-1's intangibles table for March 31, 2026 shows the following finite-lived intangible assets on the balance sheet:
| Asset Category | Useful Life | Gross Value | Net Value |
|---|---|---|---|
| User Base | 9.0 years | $1,277M | $791M |
| Advertising Customer Relationships | 5.0 years | $745M | $235M |
| Brand | 5.0 years | $735M | $368M |
| Existing Technology | 3.0 years | $27M | $10M |
| Acquired Workforce | 2.0 years | $11M | $9M |
| Total | $2,795M | $1,413M |
The world's leading aerospace manufacturer now carries $791 million of net "User Base" on its balance sheet and $235 million of net "Advertising Customer Relationships." These are accounting artifacts of the X/Twitter consolidation working through the rocket company's financial statements, but they make the corporate identity of SpaceX-as-aerospace-company harder to define in any clean sense. Annual amortization expense from these intangibles ran $786 million in 2025, $847 million in 2024, and $738 million in 2023.
TEN THINGS IN THIS S-1 THAT MIGHT GENUINELY SURPRISE YOU
1. 1.3 Billion Performance Shares for Musk
The S-1 reveals two performance grants totaling approximately 1.3 billion Class B shares. Vesting requires up to $7.5 trillion market cap AND a permanent 1-million-person Mars colony AND orbital data centers producing 100 terawatts of compute. This is, in absolute terms, the largest performance-based executive compensation package in the history of publicly traded companies. And the targets are so remote and imprecise that they function more like wish lists than binding commitments.
2. Selling Compute to Anthropic for $45 Billion
Already covered above, but worth restating because I don't think it gets enough attention. The IPO prospectus shows investors a massive revenue stream coming from the direct competitor of the AI division they just acquired.
3. Tesla Put $2 Billion Into xAI Pre-Merger
Already covered above. This is the single most governance-concerning disclosure in the entire filing.
4. Musk Can Only Be Fired By Himself
The S-1 states Musk "can only be removed from our board or these positions by the vote of Class B holders" — the same super-voting shares he himself controls. CalPERS, NYC, and NYS Comptrollers issued a joint letter demanding this be removed. It wasn't.
5. One Booster Has Flown 34 Times
The most-flown Falcon 9 first-stage booster has been launched and recovered 34 times. The original target when reusability was introduced was 10 flights per booster. This is not just impressive; it represents an operational discipline that has no real parallel in industrial history.
6. SpaceX Bills Tesla For Musk's Private Jet
Covered above. $2 million in 2025.
7. They Don't Insure Their Rockets
From the risk factors: *"We do not typically obtain insurance coverage for our satellites, payloads, or launch vehicles, and as a result we bear the full financial cost of any such losses."* The company accepts 100% of launch failure risk. This is a deliberate strategic choice, not negligence — insurance for launches like this would be astronomically expensive and the underwriters would price it to reflect the catastrophic nature of the risk. SpaceX is essentially saying: *"We're confident enough in our 98%+ success rate that self-insurance is cheaper than buying coverage."* That is a bold bet. It has paid off so far.
8. xAI Bought $506M of Megapacks From Tesla
Covered above. Tesla manufacturing → xAI data centers, all within Musk's controlled ecosystem.
9. Four Classes of Common Stock
SpaceX has authorized Class A, Class B, Class C, and Class D common stock. Class D has an unusual par value of $0.0001 per share, distinct from the $0.001 par value of the other three classes. The existence of four stock classes — not the familiar two — adds another layer of structural complexity that almost no one outside the filing has noticed.
10. $791 Million of "User Base" on the Balance Sheet
The rocket company now carries a $791 million net intangible asset called "User Base" — an accounting artifact of absorbing X/Twitter through the xAI consolidation. Plus $235M of "Advertising Customer Relationships" and $368M of "Brand." This is what happens when you merge a social media platform into an aerospace company's balance sheet: you get a corporate identity crisis in accounting form.
THE $28.5 TRILLION TAM CLAIM
Every S-1 needs a total addressable market section, and SpaceX's is an all-timer. The company claims a quantifiable TAM of $28.5 trillion, and explicitly notes that the figure excludes China and Russia.
| Segment | TAM Estimate | Notes |
|---|---|---|
| Space | $370B | Space-enabled solutions |
| Starlink Broadband | $870B | Global fixed broadband |
| Starlink Mobile | $740B | Satellite-to-mobile services |
| AI Infrastructure | $2.4T | Data center compute |
| AI Consumer Subscriptions | $760B | Grok and similar |
| AI Advertising | $600B | Digital ad market |
| AI Enterprise Applications | $22.7T | Macrohard, agentic workflows |
| Total | $28.5T | Excluding China and Russia |
For comparison, global GDP in 2024 was approximately $110 trillion. SpaceX is claiming a serviceable addressable market equivalent to roughly 26% of all global economic output.
The $22.7 trillion figure for "AI Enterprise Applications" is doing most of the heavy lifting. That number is based on the theory that AI agents will replace large portions of white-collar labor, and SpaceX (through Macrohard, the joint venture with Tesla) will capture a meaningful slice of that displacement.
Make of that what you will.
STARLINK: THE ACTUAL BUSINESS (AND THE WORRYING TREND)
Let me zoom in on Starlink specifically, because it's the only part of this company that generates real, durable cash flow.
Starlink ARPU (average revenue per user, per month) has been falling fast:
| Period | ARPU |
|---|---|
| 2023 | $99/mo |
| 2024 | $91/mo |
| 2025 | $81/mo |
| Q1 2026 | $66/mo |
Starlink ARPU has declined 33% in three years. The bull case is that subscriber growth has more than offset the ARPU decline (revenue is still growing rapidly at 50%+), and the falling ARPU reflects Starlink expanding into lower-income emerging markets and adding lower-priced tiers. The bear case is that Starlink is now competing against terrestrial broadband in markets it didn't used to compete in, and pricing pressure is real and getting worse.
The S-1 doesn't editorialize, but my read is that Starlink is trading dollars for scale. They want to lock in as many customers as possible before competing constellations — Amazon's Kuiper, China's Guo Wang and SatNet — get to commercial scale. This is a classic land-grab strategy: get massive user density first, raise prices (or reduce subsidies) later.
GOVERNANCE: THE STRUCTURE THAT PROTECTS MUSK FOREVER
The S-1 states clearly: *"Following the completion of this offering, we will have two classes of common stock issued and outstanding: Class A common stock and Class B common stock."* But then it goes further — there are also Class C and Class D shares authorized, bringing the total to four classes.
Each share of Class A common stock will entitle its holder to one vote per share. Each share of Class B common stock will entitle its holder to 10 votes per share.
The S-1 further states that Musk "can only be removed from our board or these positions by the vote of Class B holders" — the super-voting shares he himself controls. As a result, Mr. Musk will be able to control the outcome of matters requiring shareholder approval, including:
- The election of a majority of the board (through Class B shares)
- The remainder of the board (through combined voting power)
- All amendments to the charter
- All mergers and acquisitions
As a result, the company will be a "controlled company" under Nasdaq rules and intends to rely on exemptions from certain corporate governance requirements, including the requirement to have a majority of independent directors and an fully independent compensation committee.
CalPERS, NYC Comptroller, and NYS Comptroller issued a joint letter on May 13, 2026 demanding that these provisions be changed. They were not.
THE VALUATION PROBLEM
SpaceX is targeting a valuation of $1.75 trillion to $2 trillion for what will likely be the largest IPO in history. The company is raising approximately $75 billion.
At a $1.75 trillion valuation against $18.7 billion in trailing revenue, that's roughly 94x revenue. Even for a high-growth company, this is extraordinary. Let's compare:
| Company | Revenue (TTM) | Market Cap | Rev Multiple |
|---|---|---|---|
| Microsoft | ~$245B | ~$3.4T | 13.9x |
| NVIDIA | ~$130B | ~$3.3T | 25x |
| Meta | ~$164B | ~$1.5T | 9x |
| Amazon | ~$639B | ~$2.2T | 3.4x |
| SpaceX (target) | $18.7B | $1,750B | 94x |
SpaceX would be the most expensive IPO relative to revenue in modern history by a factor of 3-4x over even the most exuberant tech valuations we've seen.
Now, SpaceX's 34% CAGR is better than most of these companies were at similar points. Starlink's growth trajectory is real. But the valuation assumes perfection across every business line — Starlink must keep growing, AI infrastructure must monetize at the scale promised by the $22.7T TAM, xAI must become a profitable enterprise, and all of this must happen while SpaceX continues funding Mars and orbital ambitions.
The risk/reward at $1.75T is not obviously attractive. It prices in a future where SpaceX is not just the leading aerospace company but the leading AI infrastructure company, the leading global satellite broadband provider, and the leading enterprise AI platform — all simultaneously. That is not impossible, but it requires the company to execute at a level that has no precedent.
THE LAUNCH NUMBERS: STILL THE MOST IMPRESSIVE PART
For all the talk about AI and satellites and IPO mechanics, the foundational fact about SpaceX is still that it launches rockets at a cadence no one else can touch.
The most-flown Falcon 9 first-stage booster has been launched and recovered 34 times. The original target when reusability was introduced was 10 flights per booster. This is not just impressive engineering; it represents an operational discipline and a cost structure that creates a competitive moat no other aerospace company can cross.
SpaceX has conducted over 700 launches as of mid-2026. No other company in the world has come close to 200. The reusability economics — a refurbished booster costs a fraction of a new one — mean that SpaceX's marginal cost of launch continues to fall while everyone else's stays flat. This is a structural advantage, not a temporary one.
If you strip away the AI drama, the governance concerns, and the surreal intangible assets, SpaceX is still the company that made space affordable. That is a real business, with real economics, and a real competitive advantage.
MY BOTTOM LINE
The SpaceX S-1 documents the emergence of something unprecedented: a company that controls both the means of access to space and a significant portion of the AI infrastructure stack, wrapped in a governance structure that makes it functionally unaccountable to public shareholders.
The core business — rockets and Starlink — is genuinely world-class. The financials, stripped of the xAI drag, show a company that was profitable before the merger and is growing at 34% annually. Starlink is a $11.4 billion broadband business at 63% margins with 10.3 million subscribers and real pricing power ahead.
But you are being asked to pay $1.75 trillion for the whole package, including the AI losses, the governance issues, the related-party transactions, the surreal intangible assets, and the four-class stock structure that gives one man absolute control.
My assessment: at IPO pricing, this is a story stock priced for perfection. The Starlink business alone could justify $500-800 billion. The remaining $1 trillion of the valuation needs AI infrastructure to work out at a scale that is extremely speculative.
The 180-day lockup expires around December. If the post-IPO price is above the offering price, and if Musk decides to exercise any of his enormous performance share grants, the selling pressure could be enormous. If he doesn't — and given his voting control, he can't really be forced to sell — then the stock becomes illiquid and concentrated in a way that's unusual even for IPOs.
This is not a "don't touch" story. SpaceX is real. The technology is real. The competitive advantage is real. But the valuation is the kind of number that gets you killed if you take it seriously as a fundamental anchor rather than a speculative premise.
Buy Starlink exposure, not SpaceX exposure, if you have to have it. The math just doesn't work at these multiples.