Anthropic is reportedly joining the mix — Microsoft moves beyond OpenAI?

It wouldn’t be the start of a new day without Microsoft talking about AI, or making some sort of AI announcement. Today is no different. What stands out, though, is this sudden shift in strategy. It makes complete sense for Microsoft, but it’s also one that Sam Altman, OpenAI’s CEO, might not be thrilled about.

Microsoft has poured billions into OpenAI, and that commitment hasn’t gone away. But now, it looks like those billions may be shared more widely as Microsoft begins to diversify the AI tools it brings into Office 365.

It’s also not the first time Microsoft has “diversified.” GitHub Copilot already offers multiple models — including ChatGPT, xAI, Gemini, and Claude — giving developers options for coding assistance.

Inside the Microsoft–Anthropic deal

Microsoft is entering a new partnership with Anthropic to bring its Claude AI models into Office 365, marking a shift away from relying solely on ChatGPT. Unlike its deal with OpenAI, Microsoft will pay Amazon Web Services to access Claude, since Anthropic’s models are hosted on AWS.

For users, this means apps like Word, Excel, and PowerPoint will now run on a mix of Claude and ChatGPT, with Microsoft likely routing tasks between the two depending on demand and workload.

The financial details haven’t been disclosed, but the move sits within Microsoft’s massive AI investment plan, which is expected to hit $200 billion in infrastructure spending by 2028. And there’s also no change to be made to the current pricing of Copilot and Office 365 with the inclusion of Anthropic’s models.

One of the main reasons for the deal comes down to performance. Claude Sonnet 4 can now generate slides, spreadsheets, and PDFs directly in chat — a capability Microsoft likely finds appealing. It has also shown clear advantages in Office-specific tasks, such as producing more visually appealing PowerPoint slides and automating Excel functions.

Microsoft and OpenAI’s uneasy partnership

Microsoft has done just about everything it can to stay ahead in the AI race. Some might see it as burning money, and in many ways, OpenAI has been the furnace. The company received a $13 billion investment from Microsoft, which gave Microsoft a 49% stake in OpenAI’s for-profit arm. That deal also meant OpenAI’s models were hosted exclusively on Azure, and those same models became deeply integrated into Microsoft’s enterprise software.

Still, I’ve always had the impression that OpenAI never truly embraced Microsoft. That feeling was hard to shake when OpenAI released the ChatGPT app on Mac before Windows, despite Microsoft’s heavy funding. OpenAI also seems to lean toward Apple’s design philosophy, even bringing Jony Ive on board to help design new products. Ive, of course, is the creative force behind the iPod, iPhone, iMac, and Apple Watch.

There was also the 2023 mishap when Altman was briefly ousted as OpenAI’s CEO. In the five days that followed, he accepted a role at Microsoft to lead a new advanced AI team, only to return as CEO of OpenAI and withdraw from Microsoft’s offer.

Of course, that’s just speculation on my part as someone watching things unfold. But it does raise questions about the dynamic between Microsoft and OpenAI. For Microsoft, the answer seems clear: it’s now looking to diversify the AI models that power its suite of enterprise tools.

Looking ahead: diversification as the new normal

Similar to how there are growing calls to build more global connectivity lines after the Red Sea cables were cut, Microsoft is also looking outward. Its recent partnership with Nebius to expand data center capacity is another sign that diversifying infrastructure is becoming the norm.

In many ways, diversification is simply a smart move — one that makes sense not just for companies, but for individuals too. Spreading risk and avoiding overreliance on a single option is a principle that works in technology as much as it does in life.

For Microsoft, this shift gives it more of an edge while it continues to develop its own internal models. It also highlights how quickly the way companies interact with AI is changing. We’re clearly not at a “one model fits all” stage of AI yet, and Microsoft seems determined to use the best model for the job instead of locking itself into one provider.