The editorial frames the $1.25B Anthropic commitment as the most strategically loaded disclosure in the S-1, arguing that Anthropic — Amazon's flagship Claude partner and a vocal differentiator from Musk-aligned AI — paying a Musk-controlled entity for capacity on an xAI-built cluster would have been unthinkable a year ago. They read this as a 'tell' about how desperate frontier model labs are for compute capacity in mid-2026.
Was the first commenter to surface the buried Cloud Services Agreement language, flagging the $1.25B Anthropic commitment for access to COLOSSUS and COLOSSUS II compute capacity as the most newsworthy line item in the filing.
Singled out the reaffirmation of the xAI/X (Twitter) consolidation as the part of the S-1 they liked least, suggesting that bundling Musk's other ventures into the SpaceX corporate orbit creates governance and related-party concerns that public investors should scrutinize.
Notes that at $18.7B in 2025 revenue, SpaceX wouldn't even crack the top 700 US companies by revenue, yet private secondaries have already pushed implied valuation past $400B with a $1B-per-percent IPO price. The implication is that public-market investors are being asked to validate a price multiple that the underlying financials don't yet support.
The editorial highlights that by the filing's own segment breakdown, the consumer and enterprise satellite broadband business is the only part of SpaceX generating meaningful free cash flow. Despite $18.7B in revenue and $6.6B adjusted EBITDA, the company still posted a $2.6B operating loss and $4.9B net loss — implying the launch business remains a cash sink that Starlink subsidizes.
SpaceX filed its S-1 ahead of an IPO that markets are pricing above a $1B-per-percent valuation, and the document contains more interesting disclosures than the rocket company's usual press releases. The headline financials: $18.7B in 2025 revenue (up from $14.0B in 2024), $6.6B adjusted EBITDA, a $2.6B operating loss, and a $4.9B net loss. Starlink is doing the heavy lifting — the consumer and enterprise satellite broadband business is, by the filing's own breakdown, the only segment generating meaningful free cash flow.
Buried in the related-party and customer sections is the disclosure that caught Hacker News commenter `impulser_` first: "in May 2026, we entered into Cloud Services Agreements with Anthropic PBC ... with respect to access to compute capacity across COLOSSUS and COLOSSUS II," with the customer committing to pay SpaceX $1.25 billion. Colossus is the xAI-built Memphis supercomputer cluster; Colossus II is its successor. The S-1 also reaffirms the xAI/X (Twitter) consolidation that several commenters, including `kentm`, flagged as the part of the filing they liked least.
The filing arrives in a market where private secondaries had already pushed SpaceX's implied valuation past $400B. Public-market investors are now being asked to validate that price against financials that, as commenter `arthurofbabylon` noted, would place SpaceX outside the top 700 US companies by revenue.
The Anthropic deal is the most strategically loaded sentence in the document. Anthropic — Amazon's flagship Claude partner and a company that has spent two years publicly differentiating itself from Musk-aligned AI — is now paying a Musk-controlled entity over a billion dollars for capacity on a cluster built by xAI, Anthropic's most direct frontier-model competitor. That is not a partnership any AI buyer would have predicted twelve months ago. It is a tell about how tight frontier compute supply actually is in mid-2026: when Anthropic's choice is "buy from xAI's neighbor" or "don't train," the rivalry stops mattering.
It also clarifies the corporate geometry Musk has been quietly assembling. SpaceX builds and launches. Starlink monetizes the bandwidth. xAI builds the models. X distributes them. And now SpaceX itself is becoming a compute landlord, leasing access to the same Colossus capacity that powers Grok. The S-1 frames this as customer diversification; in practice it is the first time a launch company has booked nine-figure revenue from selling GPU hours. The biomimicry-adjacent observation from `Jabbles` — that orbital data centers will probably never beat terrestrial ones on thermodynamics — is correct but mostly beside the point. The compute Anthropic is buying is in Memphis, not orbit. The space company is selling earthbound silicon.
The financials deserve their own scrutiny. A $4.9B net loss on $18.7B of revenue is not a software-company loss profile; it is a capital-goods loss profile, where depreciation on Starship test articles and Starlink satellite constellations consumes the EBITDA before it reaches the bottom line. `Eldodi`'s summary on HN — "Starlink seems to be a real cash machine, not as good as ads but enough to support AI bets" — is the cleanest one-line reading of the document. Roughly two-thirds of revenue now comes from Starlink subscriptions and enterprise contracts; launch services, the original business, is increasingly a cost center that exists to deploy more Starlink satellites and amortize Starship development.
The community skepticism is sharper than the analyst notes have been. `kentm`'s discomfort with the xAI/X rollup reflects a real governance concern: SpaceX shareholders are now exposed to social-media moderation liability, AI training-data lawsuits, and Grok's regulatory surface area, none of which were part of the thesis when secondaries were trading at $200B. `arthurofbabylon`'s revenue-multiple critique is the quieter problem — at the rumored IPO range, SpaceX trades at roughly 22x revenue and 60x adjusted EBITDA, multiples normally reserved for pre-revenue biotech or top-decile SaaS.
If you build on Anthropic's API, the practical takeaway is that Claude's underlying compute footprint is now multi-vendor and multi-cluster in a way the marketing pages don't advertise. Capacity is being sourced from AWS Trainium, Google TPUs, and — per the S-1 — Colossus GPUs operated by SpaceX. For most workloads this is invisible. For workloads sensitive to data residency, sovereign-cloud claims, or the specifics of where your prompts are physically processed, it is worth re-reading your DPA. Anthropic's enterprise contracts allow them to subcontract compute; this filing is the first public confirmation of how broadly they are exercising that right.
If you sell into the AI infrastructure market, the signal is starker: spot GPU capacity is now scarce enough that even direct competitors are writing billion-dollar checks to each other, and any narrative built on "we have exclusive access to NVIDIA H200s" should be discounted accordingly. The frontier labs are buying from anyone with energized racks. If your differentiation is supply, your moat depreciates with every new Colossus-class cluster that comes online.
If you're an engineer thinking about whether to buy into the IPO with your bonus, the S-1 is a useful exercise in separating the engineering story from the financial one. Starship is going to land on Mars or it isn't; that question doesn't get resolved by reading a prospectus. What the prospectus does resolve is that you are buying a satellite ISP attached to a launch business, an AI compute reseller, and a social network, at a price that assumes all four work out simultaneously.
The interesting question for the next twelve months is not whether the IPO prices well — at this float size and with this much retail demand, it almost certainly will — but whether the Anthropic deal is a one-off capacity backstop or the start of a real third-party compute business for SpaceX. If Colossus II revenue scales past the $1.25B Anthropic commitment, the company stops being a rocket-and-broadband story with an AI hedge and becomes something stranger: a vertically integrated launch-to-inference stack, with the same parent company selling rides to orbit, bandwidth from orbit, and tokens from Memphis. That is either the most interesting industrial story of the decade or the most overextended one. The S-1 doesn't tell you which.
Crazy this company will IPO for >1B with such bad financials! That said, Starlink seems to be a real cash machine, not as good as ads but enough to support AI bets.2025:- Revenue: $18.7B, up from $14.0B in 2024- Operating loss: -$2.6B- Net loss: -$4.9B- Adjusted EBITDA: $6.6B- Operating cash flow
If any company can put profitable data centers in space, it will be SpaceX. But I doubt that any company can. The difficulties of the physics and engineering of cooling seem like they will always outweigh the advantages of keeping your data center on Earth.I am annoyed by the insistence that the val
It’s surprising just how low the revenue is for SpaceX. There are some 700+ companies with larger revenue figures, and yet just a small handful exceed SpaceX’s proposed valuation.In 2026 one gets the impression that SpaceX is a huge company, among the largest in the world. It’s wild to see that its
SpaceX is incredibly exciting, but I was skeptical when XAI and Twitter were rolled into it. The S-1 here makes it even more disappointing.I did want a piece of SpaceX but the valuation here is pretty eye watering compared to the fundamentals. I don't think I can put my money into this, althoug
Top 10 dev stories every morning at 8am UTC. AI-curated. Retro terminal HTML email.
"in May 2026, we entered into Cloud Services Agreements with Anthropic PBC (“Anthropic”), 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