Major Tech Companies Reduce AI Spending as Cost Scrutiny Intensifies Before Anthropic IPO
Meta and Microsoft are significantly reducing internal usage of Anthropic's Claude AI model, signaling a shift in corporate spending priorities as companies confront mounting costs associated with AI adoption. Meta has reduced Claude Code users from approximately 60,000 to 30,000 employees, while Microsoft slashed per-employee monthly AI budgets from $100,000 to approximately $10,000 across its cloud and AI division. This pullback occurs amid Anthropic's preparations for an IPO potentially valued above $2 trillion, where analyst scrutiny on customer concentration and computing costs has intensified following reports that two unnamed customers accounted for 24% of 2025 revenue.
Anthropic's approaching public offering has exposed vulnerabilities in its business model as major clients reassess their artificial intelligence expenditures. The company's reliance on a small number of high-spending customers—with just two accounting for nearly a quarter of annual revenue—creates concentration risk that IPO investors are scrutinizing closely. Additionally, the enormous computational infrastructure required to operate Claude demands substantial capital investment, raising questions about long-term profitability at scale.
The pullback also reflects a broader corporate pattern: initial enthusiasm for AI adoption without cost discipline has given way to budget discipline and internal development. Both Meta and Microsoft are simultaneously reducing external AI spending while investing heavily in proprietary alternatives, suggesting that early customers may serve primarily as development partners rather than long-term revenue sources for startups like Anthropic.
These spending cuts could reshape the artificial intelligence industry's economics and valuation assumptions. If major technology firms successfully redirect employee usage toward internal tools, independent AI companies may face revenue pressures that could affect their ability to fund research and development. Conversely, this trend may encourage more efficient AI deployment across the industry and incentivize cost-effective model development. The outcome could influence whether AI innovation remains concentrated among well-capitalized giants or remains distributed among specialized firms.