Meta is in talks with Anthropic to lease advanced computing capabilities in a deal that could be worth up to $10 billion over two years, according to an informed source. This move could represent a shift in Meta’s strategy towards generating revenue from AI infrastructure.

The potential agreement stipulates that Anthropic, the developer of the AI assistant Claude, will make monthly payments to Meta for the duration of the contract, with either party able to terminate the deal early.

Anthropic made its proposal in June, and Meta is still considering it. The negotiations have become more complex because Meta currently does not have a dedicated business for selling its computing capabilities.

The source noted that discussions are still at an early stage and may not result in a final agreement, as reported by Reuters.

Meta’s stock fell by over 2% during Friday’s trading session before partially recovering after news of the negotiations emerged. It then settled on a slight decline in extended trading.

If the deal goes ahead, it will help Meta diversify its revenue streams away from advertising and generate income from data center infrastructure. This would also put it directly in competition with companies specializing in leasing cloud computing capabilities such as CoreWeave and Nebius, given the growing demand for AI application infrastructure.

This move comes after months of Anthropic entering into an agreement with Elon Musk’s SpaceX to utilize the full capabilities of the Colossus 1 data center in Tennessee.

Meta CEO Mark Zuckerberg stated during a shareholder meeting last May that the company's entry into the cloud computing market is "an option on the table," explaining that many companies contact Meta almost weekly to gain access to its AI models or use its excess computing power.

Bloomberg reported earlier this month that Meta is working on creating a new business to provide cloud computing services and host AI models for developers.