Meta is negotiating a deal to lease roughly $10 billion worth of artificial intelligence computing capacity to Anthropic, according to people familiar with the discussions. The arrangement would give the AI startup access to massive processing power for training and running its models, while Meta would gain a steady revenue stream from its data center investments.
The scale of the computing lease
The talks center on a multiyear agreement under which Anthropic would pay Meta for access to servers and chips housed in Meta's data centers. The $10 billion figure would make it one of the largest known computing leases in the industry. Neither company has confirmed the negotiations, and the terms could still change or fall apart.
Anthropic, best known for its Claude family of AI models, has been rapidly expanding its infrastructure. The startup has already secured cloud computing deals with Google and Amazon, but a lease with Meta would add a third major supplier and reduce its reliance on any single provider.
Prediction market sees a trillion-dollar valuation
On Polymarket, a decentralized prediction platform, traders currently put a 91.5% probability on Anthropic reaching a $1.25 trillion valuation by December 31. That implied market cap would place the company among the most valuable private tech firms in the world, rivaling the likes of OpenAI and SpaceX.
The prediction market's odds have climbed steadily over the past month, reflecting growing investor confidence in Anthropic's trajectory. The startup has raised billions from backers including Google, Salesforce, and Spark Capital, and its latest funding round valued it at around $60 billion.
Why Meta is leasing, not selling
Meta has been pouring capital into AI infrastructure, spending billions on Nvidia GPUs and custom chips. Leasing capacity to Anthropic lets Meta monetize that hardware without selling it outright, keeping the assets on its balance sheet. The arrangement also strengthens Meta's position in the AI ecosystem, giving it a direct financial stake in a leading competitor's success.
For Anthropic, the deal would provide guaranteed access to compute at a time when demand for AI chips far outstrips supply. The company has warned that compute shortages could slow its development pace, and a long-term lease with Meta would help lock in capacity for years.
Regulatory and competitive questions
The proposed lease could draw scrutiny from antitrust regulators. Meta and Anthropic are both major players in AI, and a deal that gives Meta insight into Anthropic's computing needs and model training patterns might raise concerns about competitive intelligence. The Federal Trade Commission has already signaled it is watching AI partnerships closely.
Anthropic's board will need to weigh the benefits of additional compute against the risks of deepening ties with a company that also develops its own AI models. Meta's Llama series competes directly with Anthropic's Claude, and the two have poached talent from each other in the past.
If the deal goes through, it would be the latest example of big tech companies renting out their AI infrastructure rather than selling it. Microsoft, Google, and Amazon all offer cloud computing services to AI startups, but a direct lease between two AI rivals is less common.
The talks are ongoing, and no timeline has been set for a final agreement. Anthropic's valuation, meanwhile, will be tested against the prediction market's lofty expectations as the year-end deadline approaches.
