
What happened
Microsoft cut Claude spending by more than a third, with its cloud division's monthly per-employee budget falling from $100,000 to about $10,000, and Meta halved Claude Code users to 30,000.
Why it matters
Two of the largest buyers are pulling back sharply from a tool they had spent heavily on — one 28-day stretch alone cost over $105 million at one of them.
What to watch
The pullback is partly about cost and partly about pushing competing tools, so whether the cuts deepen hinges on how quickly those in-house options are adopted.
WHO IT HITSAnthropic's sales and account teams face a concentration risk, since two of its biggest enterprise accounts are actively reducing spend. AI procurement and platform teams at large enterprises may face pressure to justify external AI tool costs against internally built alternatives.
Summaries like this, in your inbox every morning.
The pullback is not merely a budgeting exercise. Meta had already introduced AI cost-cutting measures in June, suggesting the reduction in Claude Code users was part of a broader effort rather than a sudden reversal. For Microsoft, the direction came from executives Scott Guthrie and Jay Parikh, who told employees to use in-house tools instead — a signal that the company sees external models as less central to its internal workflow.
Anthropic has grown extremely fast, and that growth is now making it a headache for incumbents. Even Microsoft has reason to worry, since tools like Claude Cowork and ChatGPT Work are increasingly turning into Office alternatives. The irony of Meta worrying about someone else using its data is hard to miss, given it reportedly wanted to restrict Anthropic's access to its training data.
The stakes for Anthropic hinge on how much of this pullback is cyclical cost management versus a permanent shift toward in-house alternatives. If the cuts are mostly about price, a cheaper offering or renewed contract could reverse them. If they reflect a strategic decision by Meta and Microsoft to favor their own tools, then Anthropic's reliance on a few major clients looks more like a structural risk.
Pick your industry and the AI tools you use, and get news related to your work every day.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.
Reflection AI launched Beam, a 501 billion-parameter open-source LLM
The AI Cybersecurity and Resilience Summit will feature a Dell Technologies-led flagship program Oct
Schneider Electric agreed to acquire software company PTC in an all-cash deal that values PTC's equity at arou…

A step-by-step guide fine-tunes Muse Glimmer, Meta's 30B vision model, locally for equation-to-LaTeX conversio…

OpenAI announced always-on agent Dots on September 30; SpaceXAI's Grok Bot and Meta Muse are pushing the same…

schema-guard, a tool by developer idk-arsh, checks AI-written SQL against a saved snapshot of real table and c…
