Anthropic is in very preliminary talks to lease computing power from Meta, according to a person familiar with the matter who spoke to CNBC. The New York Times reported the potential deal could be worth about $10 billion.
The discussions come weeks after Anthropic announced a similar arrangement with SpaceX to use the Colossus 1 data center in Memphis to improve capacity for paid Claude subscribers.
Two Major Compute Deals in Two Weeks
Anthropic’s approach to Meta marks the second time in weeks that the company has sought compute capacity from another technology company rather than building or expanding its own data centers. The SpaceX deal gave Anthropic access to Colossus 1’s Nvidia GPU clusters. A Meta arrangement would tap into Meta’s massive internal infrastructure, built primarily to serve its own AI research and products.
Access to enough AI chips remains a constraint for Anthropic, which places usage limits on its most advanced models including Fable. The company was last valued at $965 billion during its most recent funding round, and has been preparing for a potential IPO as soon as October 2026.
Meta’s Cloud Ambitions
The talks align with Meta’s own strategic shift. CEO Mark Zuckerberg said in May that the company was considering entering the cloud computing business, seeking to demonstrate to investors that its AI infrastructure spending can generate revenue beyond improvements to its existing advertising and social media products.
Meta could spend as much as $145 billion on capital expenditures, including AI infrastructure, in 2026. Last October, Zuckerberg noted that companies are regularly “asking if we have compute that they could buy from us at some premium to what we’ve bought it at.”
Dave Brown, a former senior executive at Amazon Web Services, is set to join Meta, according to CNBC, a hire that reinforces Meta’s cloud ambitions.
Compute Dependency as Industry Pattern
The pattern is clear: frontier AI labs that once aspired to control their own infrastructure are now leasing it from companies with the capital to build at scale. Anthropic’s compute partners now include Amazon (its largest investor and cloud provider), SpaceX, and potentially Meta.
For agent infrastructure specifically, this consolidation matters. Agent workloads require sustained, high-throughput compute for extended reasoning chains and tool use. The companies that control GPU allocation will increasingly shape how, when, and at what cost agents can operate.
Meta declined to comment on the talks, according to CNBC.