Partnerships

Meta Weighs Leasing AI Compute to Rival Anthropic

mm
Add Unite.AI to your preferred sources on Google

Meta is in early talks to lease artificial-intelligence computing capacity to Anthropic, a step that would turn the social network’s data-center buildout into a product for outside customers for the first time. It would also drop Meta into direct competition with Amazon (AMZN ), Microsoft (MSFT ) and Google in the cloud market. The New York Times first reported the discussions, valuing a potential deal at as much as $10 billion over two years and citing three people familiar with them.

The talks are far from a signed contract. Both companies declined to comment, the talks have been described as very preliminary, and people close to the discussions have cautioned that the reported figures are speculative. Anthropic floated the idea in June 2026 and is seeking terms that would let it pay Meta monthly and exit early, the same escape clause it wrote into a recent compute deal with SpaceX (SPCX ).

Turning a capex problem into a product

Meta has been building data centers at a pace few companies can match, and investors have grown uneasy about the bill. The company told Wall Street on its first-quarter 2026 earnings call that it expects capital spending of $125 billion to $145 billion this year, most of it for AI infrastructure and more than double what it spent the year before. Finance chief Susan Li raised the range and attributed much of the increase to higher memory prices.

All of that capacity has so far fed Meta’s own models and apps. Leasing some of it to Anthropic would let Meta recoup part of the spending and open a line of business it has never run: selling compute the way Amazon, Microsoft and Google do. The demand is not hypothetical. At Meta’s annual shareholder meeting in May 2026, CEO Mark Zuckerberg said the option was on the table:

“Almost every week there are different companies that come to us from outside asking us … if we have compute that they could buy from us at some premium to what we’ve bought it at,” he said. “We haven’t done that yet because we think that we have a use for the compute. But obviously if we get to a point where we feel that we have overbuilt, then that is an option that we have.”

The talks also land in a market where Wall Street is financing the AI buildout with a growing pile of debt, raising the pressure on Meta to show a return on its infrastructure.

Inside the economics

For Anthropic, the logic is simpler: it cannot get enough Nvidia chips. A $10 billion, two-year arrangement would work out to roughly $415 million a month, about a third of the roughly $1.25 billion a month Anthropic agreed to pay SpaceX for its Colossus supercomputers, a contract reported at around $45 billion over three years.

Chip access is the binding constraint on Anthropic’s growth. The company caps usage on its most capable models, including Fable, and has already signed multibillion-dollar compute deals with Google, Amazon, Microsoft and SpaceX to widen its supply. It is reportedly preparing for a possible initial public offering as early as October 2026, which makes locking in capacity a priority. Renting from a rival lab shows how far Anthropic will go in a market where compute stays concentrated among a handful of dominant providers.

Which chips Anthropic would actually get

Neither the report nor the companies specified what hardware Anthropic would use, and the answer matters. Meta’s fleet is a mix: Nvidia GPUs, a growing share of AMD accelerators, and more than a gigawatt of Meta’s own custom silicon co-developed with Broadcom (AVGO ), which Zuckerberg has grouped under an internal “Meta Compute” effort meant to build capacity more cheaply than rivals. Anthropic’s software already runs on Nvidia hardware, so it would most likely want Meta’s Nvidia capacity; re-tooling its workloads for Meta’s in-house chips would be a heavy lift.

Meta’s position is not one-sided. Even as it weighs selling capacity, the company is itself signing multi-year cloud deals to rent compute from others, part of a $107 billion jump in its contractual commitments last quarter. Whether the Anthropic talks produce a contract, and whether Meta opens its data centers to more customers afterward, will show whether it intends to compete with the established clouds or is simply looking to offload capacity it may have overbuilt.

Theo Nash is an AI-generated specialist at Unite.AI, covering AI infrastructure, compute, and the hardware systems that power modern artificial intelligence. His work focuses on the technical foundations behind large-scale AI workloads, including data centers, accelerators, networking, and the software stacks that tie them together.

With an analytical and engineering-driven perspective, Theo examines how advances in GPUs, custom silicon, memory architectures, and distributed systems enable new generations of AI models. He pays particular attention to performance trade-offs, energy efficiency, scalability, and the practical constraints that shape real-world deployment of AI infrastructure.

Articles authored by Theo Nash are AI-generated and reviewed by Unite.AI’s editorial team to ensure technical accuracy, clarity, and responsible coverage of the rapidly evolving AI compute landscape.