xAI Sells Compute to Rival Anthropic

xAI has agreed to provide its direct competitor, Anthropic, with massive AI computing resources in a landmark multi-billion dollar deal.

Elon Musk's xAI has entered into a significant agreement to supply large-scale AI computing power to its direct competitor, Anthropic. According to details from public IPO filings, Anthropic will purchase computing services from xAI's "Colossus" and "Colossus II" infrastructure clusters. The contract is valued at approximately $1.25 billion per month and is set to run until May 2029. This landmark deal involves two major players who are otherwise rivals in the advanced AI model and enterprise services market, making the partnership particularly noteworthy for the industry.

The agreement highlights a major strategic evolution in the AI infrastructure landscape. It demonstrates that some leading AI developers are moving away from relying exclusively on their own in-house GPU infrastructure or the services of traditional hyperscale cloud providers. Instead, they are opting to procure massive computing resources from specialized, and even competing, infrastructure operators. This trend signals the emergence of AI compute as a distinct, standalone business vertical, creating a new layer of competition and options for companies building AI. For CTOs and founders, this shift introduces new strategic considerations for sourcing the immense computational power required for training and deploying advanced models.

This deal signals a major shift in the AI infrastructure market. Specialized AI compute is emerging as a standalone business, providing an alternative to building in-house GPU clusters or relying solely on traditional hyperscalers like AWS, Google Cloud, and Azure.

Companies now have more options for sourcing large-scale AI compute, potentially from specialized providers who may offer better performance or pricing than traditional cloud giants. This could change how startups and enterprises budget for and build their AI infrastructure, creating a more competitive market for computing resources.