Meta Platforms is in early discussions with Anthropic over a potential deal to lease artificial intelligence computing capacity from Meta’s data centers in an agreement that could be worth up to $10 billion over two years, according to Reuters.
The proposed arrangement would give Anthropic access to Meta’s AI infrastructure while opening a possible new revenue stream for the Facebook parent company. Anthropic reportedly proposed the agreement in June.
Under the possible terms, the AI company would make monthly payments over two years, with both sides able to exit early, according to sources familiar with the discussions cited by Reuters and The New York Times.
A new business opportunity for Meta
The talks highlight a shift in how major technology companies are thinking about AI infrastructure. Building advanced AI systems requires enormous amounts of computing power, and companies are increasingly looking beyond their own data centers to secure enough chips and servers.
For Meta, the agreement could help justify the billions it is spending on AI infrastructure while reducing its dependence on advertising as its primary source of revenue.
Meta CEO Mark Zuckerberg has previously suggested the company could eventually sell excess computing capacity.
At the company’s shareholder meeting in May, Zuckerberg said entering cloud computing was “definitely on the table,” and added that companies were approaching Meta “almost every week” to buy access to its AI models or spare computing power, Reuters reported.
The company expects capital spending to reach as much as $145 billion in 2026, much of it focused on AI infrastructure, according to The New York Times.
Anthropic’s growing demand for computing power
The potential Meta deal reflects Anthropic’s urgent need for more AI capacity as demand for its Claude models and enterprise tools grows.
Anthropic has already signed large infrastructure agreements with other companies, including a deal with Elon Musk’s SpaceX in May to access computing power from its Colossus data center in Memphis, Tennessee. The company’s need for additional capacity shows that even leading AI developers are struggling to secure enough advanced chips and data center resources to train and operate increasingly powerful models.
AI infrastructure becomes a business of its own
If completed, the Meta-Anthropic agreement would show that AI infrastructure is becoming valuable beyond the companies that originally built it.
Traditionally, companies like Meta invested in data centers mainly to support their own services. Now, those same facilities could become commercial assets, similar to how cloud providers rent computing resources to outside customers.
However, Meta faces challenges entering a market dominated by established cloud providers and specialized AI infrastructure firms. The company does not currently operate a large-scale compute rental business, which has reportedly complicated negotiations with Anthropic.
The deal also creates an unusual relationship between two AI competitors. Meta develops its own AI models, while Anthropic builds Claude, meaning Meta could potentially become a supplier to a rival in the AI market.
What it means for the AI market
The possible agreement shows that the AI race is increasingly becoming a competition over infrastructure, not just better models.
Companies with access to chips, electricity, and large-scale data centers may gain a new advantage as AI demand continues rising. For investors, the deal could offer evidence that Meta’s massive AI investments may eventually generate direct revenue.
But until the agreement is finalized, the $10 billion figure remains only a possibility. The companies could change the terms or decide not to move forward as negotiations continue.
Also read: China’s Z.AI is reportedly approaching $1 billion in annual recurring revenue, underscoring how rapidly AI companies are scaling alongside their demand for computing infrastructure.