Meta is in talks to lease computing power to Anthropic in a potential deal worth roughly $10 billion, the New York Times reported. The agreement, if finalized, would let Anthropic tap Meta's sprawling GPU infrastructure to train and run its Claude models.
Why it matters
Meta has been building out one of the largest GPU clusters in the industry to power its own generative-AI roadmap. Renting spare capacity to a frontier lab like Anthropic would be the first major external monetization of that build-out, converting a sunk AI capex line into recurring infrastructure revenue. It would also deepen the ties between two of the most compute-hungry labs in the field.
Market impact
A $10 billion lease would be the largest hyperscaler-to-AI-lab compute agreement on record, surpassing existing cloud commitments and reshaping how the industry prices frontier-model compute. Watch for confirmation of the term length and whether the deal pulls in AWS, Google Cloud, or CoreWeave as co-funders, and whether Meta's own model roadmap absorbs capacity first.
Frequently asked questions
-
What is Meta reportedly offering Anthropic?
Meta is in talks to lease computing power from its GPU infrastructure to Anthropic in a deal that could be worth roughly $10 billion, per the New York Times.
-
Why would Meta rent GPUs to a competitor AI lab?
It would be Meta's first major external monetization of its AI infrastructure build-out, turning sunk GPU capex into recurring infrastructure revenue while deepening ties with a frontier model customer.
-
How does this compare to existing AI compute deals?
A $10 billion compute lease would be the largest hyperscaler-to-AI-lab agreement on record, surpassing existing cloud commitments and reshaping pricing for frontier-model capacity.
-
Which Anthropic models would run on Meta's GPUs?
The deal would let Anthropic use Meta's GPU clusters to train and run its Claude models, though final terms and capacity allocations have not been disclosed.
-
What should investors watch next?
Confirmation of the term length, whether AWS, Google Cloud, or CoreWeave co-fund the lease, and whether Meta's own model roadmap absorbs capacity before any goes to external customers.
CoinTelegraph