• Early May: Anthropic’s scramble for compute
• May 20: SpaceX opens its books — and leans into AI
• Financial strain and strategic bet
• May 21–22: Broader AI context and emerging risks
• Competing visions: orbital AI and a not‑so‑big behemoth
• How the deal reshapes both companies

SpaceX IPO Filing Reveals Massive Anthropic Compute Deal SpaceX’s long-anticipated IPO filing has turned a secretive rocket company into a central power broker in AI, revealing a compute deal with Anthropic so large it nearly doubles SpaceX’s annual revenue and reshapes both firms’ futures.

§Early May: Anthropic’s scramble for compute

Earlier this month, Anthropic announced it would run AI workloads on SpaceX’s new Colossus data center in Tennessee, seeking relief from a global crunch in high-end chips and data-center capacity. Within days, SpaceX’s S-1 disclosed the scale: Anthropic agreed to pay $1.25 billion per month through May 2029 for access to Colossus I and II, or about $15 billion a year — nearly double SpaceX’s 2025 revenue of $18.7 billion. Axios noted that the bill is enormous but comes as Anthropic’s revenues “are taking off” and the company is constrained mainly by lack of compute.

§May 20: SpaceX opens its books — and leans into AI

On May 20, SpaceX publicly opened its books for the first time, reporting $18.67 billion in 2025 revenue but a $4.94 billion loss, largely from AI spending. Business Insider highlighted that AI compute is now the “next trillion-dollar market” in SpaceX’s pitch, with a projected total addressable market of $28.5 trillion, $26.5 trillion of it in AI rather than space. Another analysis framed the filing as a sci‑fi manifesto, full of plans for orbital AI centers and future markets “that literally don’t exist at the moment.”

The same day, multiple outlets zeroed in on the Anthropic contract. Business Insider reported Anthropic’s payments could exceed $40 billion over the term, with both sides allowed to exit on 90 days’ notice and Anthropic receiving discounted rates during the ramp-up. Axios emphasized that the agreement “nearly doubles SpaceX’s annual revenue,” underlining its importance to the IPO story.

§Financial strain and strategic bet

TechCrunch’s reading of the S‑1 showed how dependent the new AI strategy is on such outside customers: xAI lost $6.4 billion on $3.2 billion in revenue in 2025, with AI capex running at an annualized $30.8 billion as SpaceX plans to scale Grok to “multiple trillions of parameters.” Axios concluded that SpaceX is “wildly unprofitable,” with its IPO valuation hinging on faith in future AI growth rather than current performance, noting that the Anthropic deal is one of the few visible “financial green shoots.”

§May 21–22: Broader AI context and emerging risks

Within a two‑hour window on May 21, Axios described how the Anthropic–SpaceX pact, together with Anthropic’s expected first profitable quarter, helped crystallize a new AI order built on massive infrastructure and surging demand. The Verge underscored that compute access is now so strategic that Anthropic is paying $15 billion a year for Musk’s data centers, even as its Claude models compete with SpaceX’s own Grok.

At the same time, other stories painted a more precarious picture. Business Insider detailed how Grok’s NSFW features have become an official risk factor, with SpaceX facing investigations and lawsuits over alleged nonconsensual explicit imagery, including content involving minors. Another TechCrunch piece highlighted legal and environmental risks around the gas turbines powering xAI’s data centers, now subject to an NAACP lawsuit and an EPA ruling that xAI violated federal pollution rules — even as SpaceX plans to buy $2.8 billion more turbines.

§Competing visions: orbital AI and a not‑so‑big behemoth

Beyond Colossus, Business Insider reported that the “most important thing” in SpaceX’s IPO is a plan to deploy orbital “AI compute satellites” starting as early as 2028, moving data centers into space to escape terrestrial power and permitting limits. That logic aligns with Google’s Project Suncatcher, which also envisions space‑based ML data centers.

Yet Axios argued “SpaceX is not the behemoth everyone thought,” stressing that its Starlink unit is the only profitable business and that the IPO’s lofty $1.75 trillion valuation depends heavily on speculative AI and space infrastructure bets, including the Anthropic deal.

§How the deal reshapes both companies

For Anthropic, multiple outlets cast the contract as a high‑stakes move to secure scarce GPU capacity and keep pace with rivals like OpenAI. The Axios “Two hours that changed AI” feature said the expanded SpaceX partnership — at roughly $1.25 billion per month — is central to Anthropic’s ability to sustain explosive growth and reach projected Q2 revenues of $10.9 billion and its first operating profit.

For SpaceX, the agreement helps offset enormous AI losses and validates its pitch that it can sell excess compute as a service while building toward a future of orbital data centers. The Verge noted that SpaceX is already “offering AI compute as a service at significant scale” and expects similar contracts with other AI firms.

Still, reporting from Ars Technica and others suggests that Grok itself is lagging in adoption, with SpaceX’s traditional launch and Starlink businesses now recast as ancillary to a risky AI pivot centered on infrastructure rather than models. A TNW analysis added that Grok’s stalled US federal deal and falling downloads have left SpaceX increasingly reliant on renting out the very compute it built for its own AI, including leasing the Memphis Colossus 1 cluster to Anthropic for $1.25 billion a month.

Whether the Anthropic contract marks the beginning of a durable AI‑infrastructure franchise for SpaceX or a temporary patch over deep financial and legal risks remains an open question. For now, the deal cements Elon Musk as a central gatekeeper of AI compute — and binds two rival labs together in a partnership neither can easily afford to abandon.

Continue reading https://foxvector.com