A story reported by the New York Times, picked up by Reuters, CNBC, Bloomberg and CNN, says Meta Platforms is in early discussions with Anthropic to lease computing power drawn from Meta's own data centers — a deal that could reach roughly ten billion dollars over two years. The conversation is still in its earliest stages, but it marks the first time a major social-media and AI company has been discussed as a potential supplier of AI infrastructure to another leading AI lab.
For years, AI developers have rented compute from the cloud trio — Amazon Web Services, Microsoft Azure and Google Cloud — while simultaneously competing with them on AI products. Anthropic's talks with Meta would be a different model: one AI company turning its heavy capital investment in chips and servers into a revenue stream by renting the idle capacity to a peer. Nvidia's graphics chips, the bottleneck of the entire industry, sit at the center of that bargain.
The story matters because AI compute is genuinely scarce. Building and powering a frontier-AI data center can cost tens of billions of dollars before a single model is trained, and electricity grids, cooling infrastructure and chip supply all constrain how fast capacity can grow. A reliable, bankable source of compute is becoming as valuable as the models it runs. The reported deal would give Meta a concrete answer to investor criticism that its AI spending has been a one-way cost sink, converting silicon and power into a billable asset.
It also nudges Meta closer to the cloud business. Meta does not currently run a commercial cloud, but a lease of this scale would require metering, networking, security and support infrastructure not unlike what Amazon and Google sell to customers. Industry watchers see the talks as a test case: if Meta can monetize its own silicon, others with large fleets — and even with large deficits — may follow, and the line between "AI company" and "infrastructure company" will blur further.
Knowledge takeaway: AI computing power is emerging as a tradeable commodity; a reported up-to-$10-billion Meta-Anthropic lease would turn Meta's data-center capex into revenue, put Meta in cloud-adjacent competition with Amazon, Microsoft and Google, and reflect the deeper scarcity of Nvidia chips and grid-scale power that defines the current AI era.