TL;DR
Anthropic is in early talks for a ~$10B compute deal with Meta, the NYT reported. Meta spent $145B on AI infra and is building a cloud business to monetise it.
Anthropic is in very preliminary talks to lease computing power from Meta in a deal worth about $10 billion, the New York Times reported on Friday. CNBC confirmed the early-stage discussions. The talks come weeks after Anthropic announced a $1.25 billion-per-month arrangement with SpaceX to use the Colossus 1 data centre's Nvidia GPUs.
The deal would be another sign that Anthropic, despite approaching a $1 trillion valuation, cannot build compute capacity fast enough to meet demand for Claude. Access to AI chips remains a bottleneck, and the company places usage limits on its most advanced models. Google is also renting GPUs from SpaceX for $920 million per month, and Google itself is rationing Gemini access to Meta because it cannot serve demand.
For Meta, a deal with Anthropic would mark the beginning of the cloud computing business CEO Mark Zuckerberg has been signalling. In May, Zuckerberg said Meta was considering entering cloud, and last October he noted companies “regularly ask if we have compute they could buy from us at some premium.” Meta could spend as much as $145 billion on capital expenditure in 2026. A former senior AWS executive, Dave Brown, is set to join Meta. Meta cut 8,000 jobs in May while redirecting billions toward AI infrastructure, and selling excess compute to Anthropic would help justify that spending to investors.
The 💜 of EU tech
The latest rumblings from the EU tech scene, a story from our wise ol' founder Boris, and some questionable AI art. It's free, every week, in your inbox. Sign up now!
The arrangement would create an unusual dynamic: Meta, which builds its own Llama models and competes with Claude, would simultaneously be Anthropic's infrastructure provider. But the compute shortage has made such arrangements routine. SpaceX sells GPUs to both Anthropic and Google. Google sells Gemini to Meta while rationing its access. The AI industry's competitive lines are blurring because everyone needs more compute than they can build alone.