Microsoft to add Anthropic models to Office 365 Copilot as ties with OpenAI fray, reports say

Microsoft is preparing to integrate Anthropic’s AI models into its Office 365 Copilot suite, shifting from exclusive reliance on OpenAI amid strained negotiations between the software giant and its top AI partner, The Information reported.

The move would route some Copilot features — notably advanced spreadsheet automation and presentation design — to Anthropic’s Claude Sonnet 4, while keeping OpenAI’s GPT-5 in place for other tasks, according to people involved in the project cited by The Information. The company is expected to pay Amazon Web Services for access to Anthropic’s models, Reuters reported.

Internal testing, the report said, found Anthropic’s models produced stronger results than OpenAI’s systems for generating PowerPoint decks and managing complex Excel functions, prompting Microsoft executives to broaden the roster of AI providers powering its productivity tools. The integration work has been led by Charles Lamanna, who was elevated by CEO Satya Nadella to oversee business applications including Copilot, the report said.

Microsoft’s Office 365 Copilot, sold as an add-on to the company’s widely used productivity suite, will remain priced at $30 per user per month, The Information added. The product reaches into a software ecosystem used by more than 430 million paying customers worldwide and, by some analyst estimates, the Copilot add-on could already be generating more than $1 billion annually if roughly 1% of customers subscribe.

The decision underscores a broader, pragmatic shift by Microsoft toward a multi-vendor strategy for enterprise AI, the reporting said. Earlier this year the company announced GitHub Copilot would also tap Anthropic models for advanced features. Microsoft has also explored alternatives for its consumer-facing Copilot app, which has far fewer users than ChatGPT.

The changes come against the backdrop of tense negotiations between Microsoft and OpenAI over the startup’s proposed restructuring into a public benefit corporation and the degree of equity and privileged model access Microsoft would retain. Those talks, together with regulatory scrutiny in California, have delayed OpenAI’s long-anticipated initial public offering until at least 2026, the reports said.

OpenAI’s nonprofit parent has said it intends to preserve mission-driven oversight as the company restructures, while investors including SoftBank have pressed for clarity and liquidity. OpenAI declined to comment when reached by The Information. Anthropic and Amazon Web Services did not respond to requests for comment, the report said.

A Microsoft spokesperson told The Information that OpenAI would continue to be a partner for frontier models and that the company remained committed to its long-term relationship with OpenAI. Unlike its arrangement with OpenAI, where Microsoft’s large investment has granted deep access to the startup’s models, the Anthropic integration will involve paid access through AWS, the report noted. The use of a rival cloud provider to host productivity capabilities is notable given the competitive positions of the companies involved.

Companies have been cautious in adopting AI-powered productivity features; some enterprise customers have cited bugs in AI-generated presentations and spreadsheets. The reported move to add Anthropic models reflects both a response to those performance concerns and a hedge amid uncertain commercial and governance talks with OpenAI.

Microsoft’s reliance on both OpenAI and Anthropic highlights the intensifying competition among AI startups seeking enterprise footholds while illustrating how major cloud and software vendors are balancing technical performance, commercial terms and strategic relationships in their AI product road maps.