Meta Ready to Sell Computing Power to Anthropic: $10 Billion Deal in the Works?

The artificial intelligence sector might witness an unusual collaboration between two companies that are, at least in terms of language models, direct competitors. According to reports from the New York Times and confirmed by CNBC, Anthropic has initiated preliminary talks with Meta to rent computing capacity intended for the training and execution of its AI models, in an agreement that could reach a value of around $10 billion.

There had been initial discussions about Meta's intention to explore a new potential business avenue a few weeks ago, but now concrete steps are emerging. The negotiations are still in their very early phases, and there are no guarantees that they will result in a definitive agreement. If successful, the operation would represent a significant step for both Anthropic and Meta, outlining a new model of collaboration in a sector where competition and shared infrastructure are increasingly coexisting.

For Anthropic, the agreement would highlight once again how the availability of GPUs is today one of the main bottlenecks in the entire industry. The company continues to witness growing demand for Claude and its more advanced models, so much so that it has imposed usage limits on certain functionalities. The need to rapidly expand computing capacity recently led Anthropic to strike a deal with SpaceXAI to leverage the power of the Colossus 1 data center, based on NVIDIA GPUs, with a financial commitment of about $1.25 billion per month.

This strategy confirms how major AI developers are increasingly willing to turn to external infrastructures to ensure access to the chips necessary for training and inference of models. Other companies in the sector are also following similar paths: Google, for instance, utilizes computational capacity provided by SpaceX and continues to manage the availability of resources dedicated to its AI services carefully.

For Meta, this potential agreement would represent a first concrete step towards a commercial cloud computing activity dedicated to artificial intelligence. Meta has heavily invested in infrastructure and, in the current year alone, estimates suggest it could allocate up to $145 billion in capital expenditures, a significant portion of which is dedicated to building infrastructure for artificial intelligence.

Monetizing part of this capacity through cloud services would allow the company to financially capitalize on unprecedented investments and provide investors with new revenue prospects beyond the traditional advertising business. This direction is further strengthened by the anticipated entry into Meta of Dave Brown, a former senior executive at Amazon Web Services, who has extensive experience in the cloud computing market.