Meta internally projected that it could spend as much as $10 billion a year on Anthropic's AI models, according to a New York Times report summarized by Techmeme. The same report says the projection existed even as Mark Zuckerberg publicly criticized Anthropic. The provided material does not include the content of Zuckerberg's criticism, the timing of the projection, or whether Meta ultimately committed to any spending. That makes this a notable but still developing account. The $10 billion figure is framed as an internal projection of what Meta could spend, not as a signed contract, actual run-rate, or completed deal. The source summary also does not specify which Anthropic models Meta considered using or for what products. For now, the supported takeaway is narrow: Meta reportedly evaluated a very large potential annual spend on a rival AI model provider, while its chief executive was publicly taking aim at that same company. The scale of the projected spend, if accurately reported, underscores how expensive frontier model access can become for large platforms building AI products. Who benefits: Anthropic would benefit if any such projected spending became actual customer revenue. Meta could benefit if external model access helped support AI products or internal needs, but the source summary does not specify the use case. Who's exposed: Meta is exposed to scrutiny over the gap between reported internal planning and public criticism. The report also leaves open questions about how dependent large AI platforms may be on outside model providers.