Meta may be paying billions a year to the AI lab its CEO keeps publicly running down.
Meta has become one of Anthropic's largest customers even as Mark Zuckerberg criticizes the company in public. At one point this year, Meta internally projected it could spend as much as $10 billion annually on Anthropic's models, according to a New York Times report citing two unnamed sources. That figure is a projection, not a disclosed contract, and it's not clear how much Meta has actually spent so far. Still, the number is large enough to make Anthropic one of Meta's biggest vendors of any kind.
The gap between rhetoric and spending is the real story here. A company doesn't pencil in ten figures a year for a competitor's technology unless that competitor is doing something its own models can't yet replicate at the same quality or speed. For Anthropic, quietly becoming indispensable to a public critic is a better validation of its models than any press release could offer.
Trash the competition in interviews, buy its product in bulk: it's a familiar move in tech, and rarely as contradictory as it looks from the outside.