Meta and Anthropic Strike a $10 Billion AI Compute Deal

On July 17, 2026, the New York Times reported that Meta Platforms is in talks to lease part of its AI infrastructure to Anthropic, in a deal that could be worth up to ten billion dollars over two years. The proposed agreement would make Meta a direct supplier of computing power to one of its biggest rivals in the race to build frontier AI models. Anthropic first approached Meta about the arrangement in June, after demand for its enterprise product Claude Code surged sharply. The two companies are still negotiating the final size and structure of the deal. If fully drawn, Anthropic would pay Meta roughly 417 million dollars every month, and either side could walk away before the contract ends. Meta has spent tens of billions of dollars building data centers and GPU clusters to train its own Llama models and to run recommendation systems across Facebook, Instagram, and WhatsApp. The company is also preparing to launch a custom AI chip called Iris in September, designed to lower its own inference costs and reduce dependence on Nvidia. Now Meta wants to turn that spare capacity into a new revenue stream outside of advertising. Industry analysts say the talks mark a turning point for the AI industry. For years, the largest cloud providers, including Amazon, Microsoft, and Google, have been the main suppliers of AI compute. If a top model maker starts buying capacity from another top model maker, the line between competitor and supplier begins to blur, and the cost of frontier AI may rise even faster for everyone else.