The AI company racing to IPO just signed a compute deal with the man who's been calling AI safety research a scam for years.
The Summary
- Anthropic signed a deal with SpaceX/xAI to use all compute capacity at the Colossus 1 data center, part of its infrastructure buildout ahead of a planned June IPO
- The partnership pairs Elon Musk's combined SpaceX/xAI entity with Anthropic, a company he's publicly criticized for its AI safety approach
- Anthropic is locking in compute capacity as it prepares to go public, signaling infrastructure is the bottleneck everyone's racing to solve
The Signal
Anthropic announced Wednesday it will consume the full compute capacity of Colossus 1, the data center operated by Musk's merged SpaceX and xAI operations. This marks the latest infrastructure partnership for the Claude maker as it gears up for a June IPO. The deal is notable less for its technical specs than for who's sitting across the table.
Elon Musk has spent years publicly dismissing AI safety research as overblown, calling out the "effective altruism" crowd that seeded Anthropic's founding team. Now his infrastructure is powering one of the most safety-focused AI labs in the business. The SpaceX/xAI combination gives Musk control over both satellite infrastructure and AI compute, making him a critical vendor even to competitors.
"The partnership stitches another piece onto Anthropic's AI infrastructure buildout."
The timing matters. Anthropic is reportedly targeting a June IPO, which means investor roadshows are happening now. Locking in compute capacity before going public sends a clear message: we've secured the resources to scale, and we're not dependent on a single cloud provider. The Colossus 1 deal diversifies Anthropic's compute stack beyond the usual suspects (AWS, Google Cloud, Azure). It also suggests the company expects demand for Claude to spike post-IPO and needs the infrastructure already in place.
What's less clear is the financial structure. Is Anthropic paying market rates, getting a discount, or trading equity for compute credits? The sources don't specify terms, but the deal's existence points to a broader trend: AI companies are willing to do business with ideological opponents if it solves their infrastructure problem.
Key points:
- Anthropic gets guaranteed compute capacity ahead of public markets debut
- Musk monetizes Colossus 1 infrastructure by selling to a competitor
- The deal shows compute scarcity matters more than philosophical alignment
SpaceX's role here is also worth unpacking. The company is no longer just rockets and satellites. By merging with xAI, it became an AI infrastructure play. Colossus 1 gives it a product to sell: massive-scale compute for training and inference. Anthropic becomes a showcase customer, proving the data center works at production scale. If SpaceX can sign one of the top AI labs, it can pitch others.
The Implication
Watch for more deals like this. As AI companies race to IPO or raise at higher valuations, they'll need to prove they've locked in compute. That means signing partnerships with infrastructure providers who might also be competitors. The old Silicon Valley logic of "don't do business with rivals" breaks down when the bottleneck is physical: chips, power, cooling, data centers.
For investors looking at Anthropic's June IPO, the question isn't whether Claude is good. It's whether the company has the infrastructure to keep training better models and serving millions of inference requests per day. This deal suggests they're checking that box. For Musk, it's validation that SpaceX/xAI can sell infrastructure at scale, turning a cost center into a revenue stream.