Up to $10 billion annually is what Meta Platforms could spend on Anthropic's artificial intelligence technology, the New York Times reported. The potential commitment is under consideration, per the report. At that run-rate, the annual figure moves the reported relationship into capital-commitment territory more commonly associated with large-scale infrastructure than with software licensing.
The figure is a per-year number, not a cumulative total. That framing carries weight: it implies a recurring outlay, one that compounds over whatever term is eventually attached to any agreement. No term has been reported.
The run-rate read
At $10 billion a year, AI technology begins to behave like a production input rather than a research expenditure. That is the economic category where oil, logistics capacity, and chip supply sit: costs so central to operations that securing access becomes a supply-chain question, not a software budget decision. Procurement at this scale tends to concentrate supply around a small number of capable providers, a structural outcome that has historically rewarded incumbents with pricing power. It is the same capital gravity that pulled semiconductor fabrication toward a handful of foundries over the prior decades.
Meta's reported position points to a specific internal calculation. Sourcing AI capability from Anthropic at the reported scale is a decision a company reaches when building equivalent systems in-house costs more, carries greater execution risk, or takes longer to return value. At the scale reported, the choice signals where Meta expects to source frontier AI capability for the coming years. The geopolitical dimension is not incidental: AI supply chains, like chip supply chains before them, have entered industrial policy conversations that extend well beyond corporate strategy.
For Anthropic, whose technology sits at the center of the reported talks, an annual commitment of this size restructures the revenue picture. Recurring annual payments from a large corporate buyer mean contracted throughput rather than periodic transactions. How the business plans compute investment and its own capital needs looks different under that structure.
The New York Times reported the potential spending without specifying a deal structure, a timeline for a final decision, or a term length for the commitment.