SpaceX IPO Filing Reveals $15 Billion Annual Compute Deal with Anthropic
- How the deal emerged
- A two-hour window that reframed AI
- SpaceX’s upside and risk
- Anthropic’s dependence — and escape hatch
- Competing narratives around Grok and growth
SpaceX IPO Filing Reveals $15 Billion Annual Compute Deal with Anthropic SpaceX’s long-awaited IPO paperwork has turned a behind-the-scenes compute contract into a central flashpoint in the AI race, revealing just how far both the rocket company and Anthropic are willing to go to secure the infrastructure that powers cutting-edge models.
How the deal emerged
Earlier this month, Anthropic and SpaceX quietly signed a massive compute agreement granting the AI lab access to SpaceX’s Colossus data centers in Tennessee. The companies initially disclosed only that Anthropic would tap the Colossus 1 facility, later expanding to Colossus 2 as demand grew.
On May 20, SpaceX’s S‑1 IPO filing revealed the scale: Anthropic will pay $1.25 billion per month through May 2029 — $15 billion a year — for AI compute capacity at Colossus I and II. That figure is nearly double SpaceX’s entire 2025 revenue of $18.7 billion, instantly recasting the company as a budding AI infrastructure giant rather than just a launch provider.
A two-hour window that reframed AI
The next day, reporters framed the filing as part of “two hours that changed AI,” a news cycle that showcased smarter models, surging revenues and “staggering infrastructure demands” across the sector. In this context, Anthropic’s commitment — on track for its first profitable quarter with revenue expected to double to $10.9 billion in Q2 — underscored how access to compute is now as strategic as the models themselves.
SpaceX’s upside and risk
For SpaceX, the Anthropic contract “nearly doubles” annual revenue and helps monetize spare capacity in its GPU-heavy data centers. The S‑1 describes AI compute as a way to “monetize unused compute capacity” while retaining the option to redirect resources to internal projects, including its own Grok models.
But the filing also reveals deep losses: SpaceX’s AI division lost $6.3 billion on $3.2 billion in revenue in 2025, after spending $12.7 billion on AI capex that year and another $7.7 billion in Q1 2026 alone. The company is even weighing “manufacturing our own GPUs,” a move that would pit it against entrenched chipmaker Nvidia.
Anthropic’s dependence — and escape hatch
For Anthropic, the deal reflects both rapid growth and acute scarcity. The lab has been “hampered by a lack of compute power,” making Colossus capacity critical to serving its booming customer base. An Anthropic spokesperson confirmed the $1.25 billion monthly tally, noting the compute will be used primarily for inference — running models at scale rather than training them.
Yet both sides have preserved flexibility: either company can terminate the agreement with 90 days’ notice, a clause observers say is “necessary based on the fast-moving nature of the AI industry.” That matters because Anthropic’s Claude directly competes with xAI’s Grok, which relies on the same SpaceX-controlled infrastructure.
Competing narratives around Grok and growth
While the Anthropic contract strengthens the IPO story, SpaceX’s in-house AI product is under pressure. Downloads of Grok have fallen from 20 million in January to 8.3 million in April, and its paid conversion rate is reportedly just one-fifth of ChatGPT’s. A low-priced federal contract — “the $0.42-per-agency GSA deal” — is stalled, even as SpaceX rents out its Memphis Colossus 1 cluster to Anthropic for $1.25 billion a month.
Still, Elon Musk is using the moment to burnish SpaceX’s broader narrative. Sharing praise for the company’s achievements in its IPO filing, he amplified a description of SpaceX as “such a bad ass company” and added, “The SpaceX team is incredible!” on X. In another post, he highlighted a friend’s account of a SpaceX engineer happily working late into the night, “energized” by a sense of shaping “a multi-planetary future for humanity.”
As SpaceX markets itself as both launch leader and AI backbone, and Anthropic races toward profitability on the back of record compute spending, the 90-day exit clause looms over a partnership that could reshape — or rapidly realign — the balance of power in the AI infrastructure boom.
Continue reading https://foxvector.com
Write a comment