Meta Spending on Anthropic AI Models Hits Hundreds of Millions

Meta is currently spending hundreds of millions of dollars per month on Anthropic’s artificial intelligence models, even as CEO Mark Zuckerberg publicly warns against the consolidation of power among leading AI labs [1]. The social media giant has become one of Anthropic’s largest customers, with internal projections at one point this year estimating annual spending could reach US$10 billion [1].

The tension between reliance and rivalry

The financial relationship between the two companies highlights the complex "friend-foe" dynamics defining the current AI race [1]. While Meta develops its own internal tools, it has relied on Anthropic’s models to power and test "Hatch," an upcoming personal AI agent designed to assist users with health, finances, and relationships [1].

Zuckerberg has publicly criticized leading AI labs for painting a future "filled with doom" and consolidating power, though he did not name Anthropic directly [1]. Despite this rhetoric, Meta’s leadership is acutely aware that their spending levels directly influence Anthropic’s financial trajectory as the startup prepares for a potential initial public offering [1]. Meta’s head of AI product, Nat Friedman, has signaled to employees that shifting usage toward internal coding tools or OpenAI’s products could lower Anthropic’s revenue ahead of that IPO [1].

Competitive infrastructure and efficiency

Anthropic is positioning itself as a leaner alternative to rivals like OpenAI by focusing on enterprise demand and lower compute costs [2]. The company projects its compute spending will reach US$6 billion in 2025, compared to US$15 billion for OpenAI, with that gap expected to widen significantly by 2028 [2]. Anthropic aims to become cash-flow positive by 2027, two years ahead of its competitor’s projected timeline [2].

This efficiency strategy is supported by a revenue model where 80% of income is derived from paid API usage, contrasting with the free-heavy user base of services like ChatGPT [2]. Meanwhile, the broader market for AI infrastructure remains aggressive, with Microsoft and Google recently pledging over US$16 billion for projects in Europe [2].

What to watch

• Meta’s internal shift: Monitor whether Meta continues to migrate its internal projects, such as Hatch, away from Anthropic’s models to its own proprietary tools [1].
• Anthropic’s IPO timeline: Watch for further reports on the startup’s valuation and revenue growth as it approaches its target of becoming cash-flow positive in 2027 [1, 2].
• OpenAI’s response: Observe how OpenAI’s updated coding tools, such as the revamped Codex, impact the market share of Anthropic’s Claude Code, which is currently favored by many businesses [3].

The extent of Meta’s financial entanglement with Anthropic remains a significant, previously unreported factor in the AI sector. Whether Meta continues to fund its rival or pivots to fully internalize its AI infrastructure will likely determine the competitive landscape for both firms leading up to Anthropic’s anticipated public market debut [1].