Microsoft cuts internal Claude usage spending by one-third

Two people familiar with the spending said Microsoft executives estimated earlier this year that the company's internal use of Anthropic technology would cost at least $1 billion annually. One of the people said Microsoft subsequently asked employees to cut back on Claude usage to save costs and rely more on Microsoft's own AI tools, and the related budget was reduced by more than a third. People with direct knowledge said executives including Scott Guthrie, who heads Microsoft's Azure cloud business, and Jay Parikh, who leads the developer tools division, have asked employees to use models cautiously in code development scenarios, urged managers to monitor employee consumption, and prioritized non-Anthropic models. A current Microsoft employee said executives can now see employees' daily AI consumption as well as the specific models employees invoke on GitHub Copilot. Microsoft recently further lowered employees' overall AI usage quotas. For example, the Cloud and AI division, which has more than 60,000 employees, previously allowed each employee to use $100,000 worth of AI models per month. People familiar with the matter said the cloud division recently cut the individual monthly budget to about $10,000 in most scenarios. Current employees said the budget cuts have sparked discontent among some engineers, who complain that model usage is no longer unrestricted.