Anthropic pays xAI $1.25 billion monthly for compute access in massive AI infrastructure deal

Anthropic shocked the AI industry earlier this month when it struck a deal to buy 300 megawatts’ worth of compute power from xAI, securing the entire output of the Colossus 1 data center near Memphis, Tennessee. Now we know the staggering price tag: $1.25 billion per month through May 2029.

The massive financial commitment highlights just how expensive it has become to compete in the AI race. The deal could generate over $40 billion in revenue for Elon Musk’s xAI, making it one of the largest cloud computing contracts ever signed.

Details of the transaction emerged from SpaceX’s S-1 filing with the SEC, as reported by TechCrunch. The deal allows xAI to “monetize unused compute capacity in our infrastructure,” according to the filing. Anthropic will get a discounted rate for the first two months as xAI completes its ramp-up operations.

The arrangement puts xAI in an unusual position within the AI market. Most companies either build data centers for their own use or construct them to serve other customers – rarely both at once. This emerging “neocloud” model allows AI companies to offset massive infrastructure investments by selling excess capacity to competitors when their own usage drops.

For xAI, the timing suggests strategic necessity as much as business opportunity. Usage of Grok, the company’s flagship AI assistant, has declined significantly in recent months. This drop freed up servers that xAI now needs to monetize, particularly as the company prepares for a public offering.

The contract terms offer flexibility for both parties:

• Either company can terminate with 90 days’ notice
• xAI plans to pursue similar deals with other AI companies
• The arrangement allows multiple revenue streams from the same infrastructure

SpaceX frames the deal as smart resource management, stating that “our dual monetization strategy provides multiple pathways to generate returns on invested capital.” But the subtext reveals a common challenge in the AI industry: companies are building massive compute capacity faster than they can fully utilize it.

The deal also underscores the enormous financial barriers facing AI companies today. At $1.25 billion monthly, Anthropic’s compute costs alone exceed the entire annual revenue of many tech companies. This spending level reflects the intense competition to train and run increasingly sophisticated AI models.

For the broader AI industry, the arrangement signals a potential shift toward more flexible infrastructure sharing. As compute costs continue climbing, expect more companies to explore similar hybrid models that balance competitive positioning with financial reality.