Anthropic to Pay xAI $1.25 Billion Monthly for Colossus Compute

Details buried in SpaceX's IPO filing reveal that Anthropic will pay xAI $1.25 billion per month through May 2029 to access the full output of the Memphis Colossus 1 data center — a deal worth more than $40 billion.

4 min readSource →
Anthropic to Pay xAI $1.25 Billion Monthly for Colossus Compute

SAN FRANCISCO — When Anthropic quietly announced it had secured 300 megawatts of compute from Elon Musk's xAI earlier this month, the AI world took notice. Now, thanks to SpaceX's S-1 filing with the SEC, the full financial scope of that arrangement is clear: Anthropic will pay xAI $1.25 billion per month through May 2029 — a contract that could deliver more than $40 billion in revenue to xAI over its lifetime.

The deal gives Anthropic exclusive access to the entire output of xAI's Colossus 1 data center near Memphis, Tennessee, a facility that houses more than 220,000 Nvidia processors drawing 300 megawatts of power and representing one of the largest concentrations of AI compute in the world.

A Landmark Deal in AI Infrastructure

The scale of the arrangement is difficult to overstate. At $1.25 billion per month, xAI would collect more revenue from this single contract than most AI startups generate in a year. Anthropic is essentially renting a purpose-built supercluster — one that xAI originally built to train and run its own Grok models — and paying a premium for the privilege.

Anthropic will receive a discounted rate for the first two months while xAI completes infrastructure ramp-up, after which full monthly billing begins. Either company may exit the arrangement with 90 days' notice, giving both sides flexibility as the AI compute market continues to evolve.

SpaceX's S-1 filing framed the deal in straightforward terms: it "allows us to monetize unused compute capacity in our infrastructure." The filing also teased additional similar arrangements: "We expect to enter into additional similar services contracts."

The Neocloud Model Takes Shape

The arrangement positions xAI as what analysts are calling a "neocloud" — an AI company that builds and operates massive compute infrastructure for its own models while simultaneously renting that capacity to competitors when utilization falls short. It is an unusual stance in an industry where most players either build for themselves or build for others.

Anthropic to Pay xAI $1.25 Billion Monthly for Colossus Compute — additional image

The deal emerged partly out of necessity. Usage of Grok, xAI's flagship AI assistant, has declined in recent months as rival models from Anthropic, OpenAI, and Google have sharpened their capabilities. With servers sitting underutilized at Colossus 1, xAI found that one of its closest competitors was also its best potential customer.

"We believe our dual monetization strategy provides multiple pathways to generate returns on invested capital," SpaceX wrote in its prospectus.

Strategic Timing Ahead of SpaceX IPO

The timing of the disclosure is significant. SpaceX revealed the deal in its S-1 as part of a broader case for the company's financial health ahead of a planned public offering. xAI's infrastructure is a key asset described in the filing, and demonstrating that it generates substantial third-party revenue strengthens the investment thesis.

For xAI, the deal also buys time. With $1.25 billion in monthly guaranteed revenue flowing in, the company has a cushion to continue developing next-generation Grok models — including a version reportedly being built at trillions of parameters — without relying solely on consumer or enterprise subscriptions.

Colossus as a National AI Asset

The Colossus 1 facility has attracted attention since its construction in Memphis. The site drew scrutiny over the use of natural gas generators during its power ramp-up, but it has since secured grid connections and continued expanding. With 300 megawatts now fully committed to Anthropic's workloads, xAI has already begun planning Colossus 2.

The arrangement underscores a broader dynamic in AI: compute is the new oil. Companies that can build and operate at scale — and fill their capacity reliably — hold structural advantages regardless of which foundation model happens to be winning benchmarks in any given quarter. xAI, with one of the world's largest contiguous GPU clusters, has built exactly that kind of leverage.