Microsoft on Track to Pay Anthropic $2B+ a Year for Copilot

Microsoft's new Office chief Charles Lamanna pushed CEO Satya Nadella to start paying for Anthropic models to power Copilot, and the company is now "on track to pay Anthropic more than $2B per year," a source told The Information.

Lamanna made the pitch while Microsoft was fielding customer complaints that Copilot was, in the outlet's words, "buggy and not good enough at automating work in Excel or PowerPoint." Anthropic's latest models performed better at advanced tasks like automating financial functions in Excel and generating PowerPoint presentations from user instructions, according to the reporting, and Claude is now being blended with OpenAI technology inside Word, Excel, Outlook and PowerPoint.

He had to argue past internal concerns about the higher cost of Anthropic models, and is still "weighing Anthropic and OpenAI models to improve capabilities and margins" as Copilot competes with Claude and ChatGPT directly. Several senior executives now name him as a potential successor to Nadella.

Unanswered in the article: what exactly Microsoft is buying from Anthropic (API inference, dedicated capacity, or a longer-term commitment), and who absorbs the cost gap between the two providers. The $2B+ figure sits with one outlet and one source.

Nadella himself told enterprises this week not to trust "black box" super-intelligence systems without deterministic guards. Internally, the Copilot team is already shopping models on output quality.