Meta Reduces AI Spending on Rival Anthropic

Meta Platforms Inc. is reducing its financial reliance on AI startup Anthropic as it shifts toward internal development. The social networking company previously projected annual spending on Anthropic's tools could reach $10 billion, utilizing resources like the Claude Code tool to power and test its upcoming AI agent, Hatch.

Meta is now transitioning to its own internal tools, including Muse Code and a new AI model internally called Watermelon. This shift comes despite a complex relationship where Meta CEO Mark Zuckerberg has publicly criticized leading AI labs for attempting to consolidate power.

In a potential reversal of the financial flow, Anthropic has approached Meta to purchase up to $10 billion in computing power from Meta's data centers over a two-year period, though no agreement has been finalized. Meta executives noted that the company's decision to move away from Anthropic's tools could negatively impact the startup's revenue as it prepares for a projected initial public offering.