Meta Projected $10B Annual Spend on Anthropic's AI — Despite Zuckerberg's Public Criticism
Summary
- • Meta internally projected spending up to $10B annually on Anthropic's AI models, per NYT
- • Meta uses Anthropic's Claude Code to build its upcoming 'Hatch' AI agent internally
- • Meta pays hundreds of millions per month to Anthropic despite Zuckerberg's public criticism
- • Anthropic's potential IPO could value it at $2T; Meta's spend could influence that outcome
Details
Meta projected up to $10B annual spend on Anthropic AI models
According to a New York Times report (August 27, 2026), Meta internally projected it could spend as much as $10 billion annually on Anthropic's AI models, potentially making it one of Anthropic's largest customers.
Meta uses Claude Code to build its 'Hatch' AI agent product internally
Meta has been using Anthropic's Claude Code tool for internal development and testing of 'Hatch,' its upcoming AI agent product — revealing a deep reliance on a competitor's AI tools to build Meta's own AI product roadmap.
Meta currently paying hundreds of millions per month to Anthropic
While Meta has reduced some spending from peak projections, it continues to pay hundreds of millions of dollars per month for Anthropic's tools, according to sources cited by the NYT and subsequent corroborating reporting.
Irony: Zuckerberg publicly criticized Anthropic while Meta was a major customer
Meta CEO Mark Zuckerberg has publicly criticized Anthropic and other AI labs over industry concentration risks, even as internal documents show Meta projected becoming one of Anthropic's largest customers — a stark contradiction between competitive rhetoric and operational dependency.
Anthropic projected $65B+ annual revenue; IPO could value it at $2 trillion
Anthropic estimated in July 2026 that its annual revenue could exceed $65 billion. Meta executives were reportedly aware that their spending levels could influence Anthropic's trajectory ahead of a potential IPO that might value the startup at $2 trillion.
Details compiled via Grok live web research from multiple corroborating outlets
Information sourced from The New York Times (primary), corroborated by The Business Times, Odaily, Phemex News, and The News International via Grok live web research on August 27, 2026.
Primary source: New York Times (August 27, 2026). Corroboration: The Business Times, Odaily, Phemex News, The News International. Compiled via Grok live web research.
What This Means
This story reframes the Meta-Anthropic relationship from rivals to deeply symbiotic — Meta is simultaneously one of Anthropic's most vocal public critics and potentially its largest enterprise customer. A $10B annual spend projection would be extraordinary even for Meta's scale, and the reliance on Claude Code to build Meta's own AI agent 'Hatch' suggests the dependency is strategic, not incidental. With Anthropic eyeing a $2T IPO valuation and $65B+ in projected revenue, Meta's spending decisions carry real market influence. The competitive tensions this creates — a major customer whose own product roadmap directly competes with the vendor it funds — are likely to intensify as both companies push deeper into the AI agent market.
Sentiment
Widespread highlighting of hypocrisy and strategic contradiction at Meta
“Zuckerberg spent months publicly trashing Anthropic's models. Meta's internal teams were quietly projecting $10B a year to buy them anyway. That's not a strategy gap - that's the whole 'we build our own AI' story falling apart from the inside.”
“The hypocrisy is staggering. Mark Zuckerberg publicly slammed 'leading AI labs' for trying to 'concentrate power'... Behind the scenes? Meta is one of Anthropic's largest customers. Internal estimates show Meta could spend $10B annually on Anthropic's models.”
“Meta is spending up to $10B a year on Anthropic’s Claude while publicly racing to beat them. This is the AI industry in a nutshell: competitors bankrolling each other because their own models still aren’t good enough.”
“SITUATION EXPLAINED: Meta is one of Anthropic's largest customers while Zuckerberg attacks it publicly. Meta internally projected spending as much as $10 billion a year on Anthropic's models... Meta uses Anthropic's models to power and test Hatch.”
“$META modeled spending as much as $10B on Anthropic this year. The company Zuckerberg is racing to beat is also the one he can't stop paying. You compete with the model in public and rent it in private.”
Split
~85/15 on the irony/dependency being a major story (most reactions call out the contradiction; few defend the dual-track approach)