Meta internally assessed this year that it could spend as much as $10 billion annually on Anthropic’s AI models, potentially making it one of the startup’s largest customers, according to a New York Times report cited on August 28. People familiar with the matter said Meta continues to pay Anthropic hundreds of millions of dollars a month for access to its AI tools, despite Meta’s public criticism of concentration risks among AI labs. Meta previously used Anthropic’s Claude Code tool for internal development and to test its upcoming AI assistant, Hatch, but has since reduced some spending as it advances its own models and tools. Anthropic expects annual revenue to exceed $65 billion in July this year, meaning a full $10 billion procurement plan from Meta would make Meta a major revenue source. The relationship reflects the AI sector’s increasingly intertwined pattern of cooperation and competition.