Meta and Anthropic are negotiating a $10 billion deal for Anthropic to lease massive computing capacity to the social media giant, according to people familiar with the discussions. The arrangement would give Meta access to Anthropic's cloud infrastructure for training and running advanced artificial intelligence models.
The Deal's Structure
The proposed lease would span multiple years and cover tens of thousands of specialized AI chips, likely including Anthropic's own custom processors. Meta would pay for guaranteed compute time, similar to existing agreements between AI companies and cloud providers. Neither company has commented publicly on the talks.
Valuation Surge
Anthropic's valuation is projected to reach $1.25 trillion by December 31, with a 94% probability according to a prediction market tracked by Polymarket. That would make Anthropic one of the most valuable private companies in the world, surpassing many publicly traded tech firms. The startup has raised billions from investors including Google and Salesforce.
Why Compute Matters
AI companies are racing to secure enough computing power to train ever-larger models. Meta has been investing heavily in its own AI infrastructure but still needs additional capacity. Anthropic, which develops the Claude chatbot, has built one of the largest private supercomputers. Leasing that capacity to Meta would generate revenue while helping Anthropic amortize its hardware costs.
The talks come as both companies race to secure computing power for next-generation AI models. A deal of this size would be one of the largest compute lease agreements in the industry.
