Anthropic, the AI company behind the Claude chatbot, is in early talks to acquire computing power from Meta, according to CNBC. The discussions are described as preliminary, and no agreement has been reported.
The deal would center on compute — the raw processing capacity, typically delivered through large clusters of specialized chips, that AI companies need to train and run their models. According to reporting from Cryptonews.net, the arrangement under discussion would involve Anthropic leasing computing capacity from Meta rather than buying hardware outright.
The reported size of the potential deal varies sharply across sources. Cryptonews.net frames it as a $10 million arrangement, while Briefs Finance describes talks over a $10 billion computing deal. Because these figures conflict and the talks are early, the eventual scale — if any deal happens at all — remains unclear.
Separately, according to Yahoo! Finance Canada, the investment bank Jefferies is watching for a possible Anthropic-related deal as chipmaker AMD's "Advancing AI 2026" event approaches, a sign that Wall Street is tracking where Anthropic will source its computing power.
The talks are notable because Meta is better known as a rival building its own AI models than as a supplier to competitors. If Anthropic strikes a deal, it would show how hungry AI firms are for compute — enough to buy it from companies they otherwise compete against.
Why it matters: access to computing power has become one of the biggest constraints and bargaining chips in the AI race, and who controls it increasingly shapes which companies can build the most capable systems.