According to reports, Meta and Microsoft, Anthropic's two largest enterprise customers, are now working to reduce employee use of Claude, posing a challenge to Anthropic's efforts to sustain its explosive growth.

According to people familiar with the matter, Microsoft executives earlier this year had projected that the company's internal spending on Anthropic technology would reach at least $1 billion. One of the people said Microsoft has since cut that projection by more than a third, as the company asked employees to reduce their use of Claude to save costs and shift more toward Microsoft's own AI tools. (That figure does not include costs generated by customers using Anthropic models through Microsoft tools, spending that has been growing steadily.)

According to one person familiar with the matter, the number of employees within Meta using Anthropic's coding assistant Claude Code has fallen sharply, recently dropping to about 30,000. That is below the roughly 60,000-user scale from earlier this year. The decline in users was partly due to spring layoffs, when Meta cut about 10% of its workforce (at the time, its total headcount was 78,000). But a more important factor is that Meta developed its own AI coding tools and Muse Spark series AI models, actively promoting these products to other enterprises while also encouraging internal employees to use them.

Internal data show that Muse Code, a tool built on Meta's models and designed to compete with Claude Code (Meta began testing the tool with external customers in August), has recently surpassed 6,000 employee users, while another internal-only similar tool, MetaCode, has more than 30,000 users.

A June memo showed that Meta's spending on internal use of AI tools this year is expected to reach billions of dollars. Although the memo did not specify how that spending was distributed among different companies, Claude Code was undoubtedly the most favored external coding tool at the time. Internal data show that over the most recent 28 days, Meta spent more than $105 million on Claude Code. Meta also used Claude this year when developing the consumer-facing Muse AI agent, but switched to its own models for the final product launched last month.

These two tech giants remain major users of Claude, but even a slight cutback in spending by core customers like them can create broader headwinds for Anthropic; after all, soaring AI tool costs have already forced other enterprise customers such as Cognition to seek cheaper alternatives. Due to concerns about data privacy, some large companies engaged in sensitive enterprise business (such as Booz Allen Hamilton (BAH), NVIDIA (NVDA) and Palantir Technologies) have requested new safeguards, or reduced or even stopped using Anthropic's most advanced (and most expensive) models.

These challenges come as Anthropic prepares for what could be the largest initial public offering (IPO) ever; that process requires the company to disclose more information about its business to potential investors. According to data from Anthropic's draft IPO prospectus, nearly a quarter of the company's revenue last year came from two customers. In the first half of this year, Anthropic's revenue grew sharply: according to reports, by July its annualized revenue had exceeded $65 billion, more than seven times higher than at the start of the year; during that period, the company also added many customers. Anthropic told investors that in the 12 months ended June, about 6,000 customers spent $100,000 or more on its AI software, four times the number of similar customers in all of 2025. The company also told investors that in the 12 months ended June, more than 1,000 enterprises spent more than $1 million, and more than 100 enterprises spent more than $10 million.

Meta Platforms, Inc. (META) and Microsoft (MSFT) are both competitors and customers of Anthropic. Meta CEO Mark Zuckerberg announced last week the creation of a new division called the "Meta Enterprise Platform," focused on providing AI products to enterprise customers. Meanwhile, over the past year, Anthropic's rapid revenue growth and its flood of new enterprise software products have caught Microsoft leadership off guard.

This spring, internal use of Claude within Meta surged as employees rushed to try "tokenmaxxing" (maximizing token quotas) to demonstrate their AI technical prowess and compete for rankings on a leaderboard spontaneously created by employees called "Claudeonomics." This phenomenon prompted company leadership in June to commit to stricter AI usage controls, including setting budgets, limiting employees' token consumption, and establishing a central dashboard to monitor overuse of Claude Code and other external AI tools.

Earlier this year, Microsoft employees' use of Claude also surged sharply. According to two people familiar with the matter, as of May, Microsoft employees' annual spending on Claude-related models (including Claude Code, Claude models integrated into Microsoft Copilot, and the Claude Mythos model used for cybersecurity scanning) was projected to exceed $1 billion. According to these people, rapidly rising costs from engineers' use of Claude, combined with pressure on Microsoft to strengthen its own Claude Code competitor, prompted some executives to instruct employees to reduce reliance on Anthropic products over the following months. Company leadership cut Claude Code subscriptions and instead asked employees to use Claude models through Microsoft's GitHub Copilot tool. That tool allows users to choose which AI model to use, but also includes an "automatic" mode; in that mode, most queries are directed by default to OpenAI's models. This is much cheaper for Microsoft to operate because, thanks to its commercial partnership with the AI developer, Microsoft does not have to pay to use OpenAI models.

According to people familiar with the matter, senior leaders including Scott Guthrie, who oversees Microsoft's Azure cloud business, and Jay Parikh, who oversees the developer tools division, have asked employees to be more prudent when writing code and urged managers at all levels to monitor employee usage spending while prioritizing non-Anthropic models. In addition, according to a current Microsoft employee, company executives can now view data on each employee's daily AI spending and the specific models they use through GitHub Copilot. In an August memo to employees, Parikh wrote: "Internally, moving more workloads to OpenAI models helps us get greater value from our token investment." The contents of the memo were reviewed by media outlet The Information. "This fiscal year, every division needs to manage its own token budget, so we all need to keep learning and optimizing."

Recently, Microsoft lowered the overall limit on employee use of artificial intelligence (AI). According to people familiar with the matter, in the cloud and AI division, which has more than 60,000 employees, each employee previously had a monthly budget of up to $100,000 for AI model use; the division has recently reduced that budget to about $10,000 per month in most cases. According to current employees, the cuts have sparked dissatisfaction among some internal engineers, who complain that they can no longer use these models as freely as before.

Although Microsoft has cut its internal spending on Anthropic, these figures do not include what Microsoft pays Anthropic to provide Claude-based features to its customers, which is much larger. Microsoft had planned to invest at least $2 billion annually to use Anthropic's models to support Copilot AI features used by its productivity software customers. At the same time, Microsoft has also replaced some of its own models with Claude in Copilot; according to people familiar with the matter, since July, Microsoft has begun directing complex queries in Copilot by default to OpenAI's GPT-5.6 model. However, according to one person familiar with the matter, as the number of Copilot subscribers has increased, Microsoft's payments to Anthropic for providing these features to customers have been rising steadily this year. The person said this growth has largely offset cuts in Microsoft's internal Claude-related costs, leaving Microsoft's total payments to Anthropic roughly flat compared with peak levels earlier this year.

A Microsoft spokesperson confirmed that the company recently set OpenAI's models as the default for internal AI coding, but said engineers can still choose to use Claude models. The spokesperson noted that Microsoft has been guiding employees to use its own GitHub Copilot AI coding tool rather than external tools such as Claude Code, because using its own tools helps the company improve its products. In a statement, the Microsoft spokesperson said: "All Microsoft divisions need to manage AI use prudently and efficiently in line with their own business priorities and in compliance with standard engineering and operational norms."