Anthropic Locks In $1.25B a Month on SpaceX's Colossus and Hits Profit Two Years Early

Anthropic has committed to roughly $1.25 billion a month through 2029 for access to SpaceX's Colossus supercomputing infrastructure, Axios reported on May 21. That's about $15 billion a year and more than $45 billion through the life of the contract — and it makes Elon Musk's launch company an unexpectedly central tenant of Claude's training and inference stack alongside the lab's existing footprint at Amazon and Google.

The compute deal lands in the same news cycle as a financial milestone that is, if anything, more striking. Anthropic is on track for its first profitable quarter, with Q2 revenue projected at $10.9 billion — more than double the prior period — and an estimated $559 million operating profit. The company has told board members that figure arrives roughly two years ahead of internal projections, a sharp inflection for an industry whose defining feature has been spend running miles ahead of revenue.

What changed is the bill of materials. Claude is now deeply embedded in enterprise procurement at the scale of KPMG's 276,000 staff and a parade of similar deals, and Claude Code has eaten a meaningful slice of GitHub Copilot-style developer spend. Pair that revenue with the SpaceX contract and the math finally pencils: $15B a year of compute against revenue that is annualizing past $40B and still climbing.

The bigger story may be Musk's positioning. After a courtroom defeat against OpenAI this week and a separate SpaceX-routed buyout of Cursor lined up for the days after its June 12 IPO, SpaceX is quickly becoming the compute landlord for the labs that aren't xAI. Anthropic's Colossus contract is the loudest signal yet that, in 2026, owning the gigawatts is at least as valuable as owning the model — and Musk owns a lot of the gigawatts.