Will the Microsoft-Anthropic deal leave OpenAI out in the cold?

Microsoft wasted little time last fall after reaching a deal to finalize its new relationship with OpenAI to find a new AI dance partner — Anthropic, the second most valuable AI startup in the world. Even though the relationship between Microsoft and Anthropic is only a few months old, it appears as if Microsoft sees a future with Anthropic that’s at least as valuable as the one it had with OpenAI.

Why might Anthropic become Microsoft’s most important partner? To find out, let’s take a closer look at Anthropic and what it brings to the table.

Anthropic’s beginnings 

Anthropic was founded in 2021 when seven AI researchers, including OpenAI Vice President of Research Dario Amodei, quit OpenAI because they felt the company was compromising AI safety with its full-speed-ahead-damn-the-consequences approach.

Anthropic said it tried to build its genAI chatbot Claude in a way that ensures the chatbot can “avoid toxic or discriminatory outputs, avoiding helping a human engage in illegal or unethical activities, and broadly creating an AI system that is helpful, honest, and harmless.”

The company outlined how it would do that in a set of principles it calls Claude’s Constitution. Avivah Litan, distinguished analyst at Gartner Research, told Computerworld that Anthropic’s decision to publicly outline its principles, “starts the dialogue and, more importantly, [describes] actions regarding the principles that generative AI should be trained on to keep it safe, trustworthy, and aligned with human values and the preservation of human civilization.” (Anthropic just this month released a greatly expanded version of the Constitution.)

Those are fine words and worthy goals, but the company hasn’t always been able to live up to them. A judge ruled the company illegally downloaded pirated books and used those books to train Claude without asking permission from the copyright holders. Eventually, Anthropic agreed to pay authors and publishers $1.5 billion for what it did – the largest payout ever for copyright violations. Beyond those unscrupulous training practices, Anthropic’s technology doesn’t always work as it should; last May, the company’s own researchers found that the Claude Opus 4 model resorted to blackmail 84% of the time when threatened with a shutdown.

That hasn’t stopped Anthropic from getting investments from multiple tech giants, including Amazon, which has invested $4 billion in it so far, and Google, which has so far invested $2 billion — and has promised to invest $1 billion more.

That brings us to Microsoft.

Making a big bet on Anthropic

In late October 2025, Microsoft signed a deal with OpenAI allowing OpenAI to restructure itself to pave the way for an IPO and giving Microsoft a 27% stake in the company. Microsoft also gets to keep its intellectual property rights to OpenAI technologies until 2032. And it commits OpenAI to buy $250 billion in Azure cloud services from Microsoft.

The ink was barely dry on the pact when three weeks later Microsoft made a whopper of a deal with Anthropic in which Microsoft pledged to invest $5 billion in the company. In addition, Microsoft gave business users of Microsoft Foundry access to Anthropic’s cutting-edge Claude models. Beyond that, Anthropic will buy $30 billion in Azure cloud services.

That was just the prelude. Just this month, Microsoft committed to spending $500 million annually on Anthropic’s AI models, integrating them into Copilot as well as many other of Microsoft’s productivity tools. 

The deal makes clear that Microsoft isn’t satisfied with OpenAI’s ChatGPT, the brains behind Copilot. Because of that tool’s shortcomings, Copilot had done little until recently to expand Excel’s capabilities, something at which Claude excels. For example, Claude has been found to be 15% better than ChatGPT at handling complex financial modeling and error detection. It’s also better at summarizing large collections of documents, something ChatGPT has struggled with. So now, Copilot uses Claude rather than ChatGPT for all those and other tasks. 

Claude also handles email drafting and meeting scheduling in Copilot. In addition, Microsoft 365 Copilot uses Claude for multi-step analysis and other tasks. 

All this is part of Microsoft’s best-of-breed approach to Copilot. Rather than relying solely on ChatGPT, Microsoft is using other AI models when those models are superior.

In addition, Microsoft has said in a blog post it is now using Claude in Microsoft Foundry “to bring advanced reasoning, agentic workflows, and model intelligence purpose built for healthcare and life sciences industries.” 

That’s particularly important because Microsoft says its AI future isn’t in general purpose AI tools like Copilot and ChatGPT, but instead in highly targeted tools that solve specific problems — exactly like the ones Claude will perform in Microsoft Foundry for the healthcare and life sciences industries.

Microsoft’s deals with Anthropic make it clear that Microsoft believes its AI future is not solely tied to OpenAI. Microsoft got exactly what it wanted out of OpenAI — a headstart in the AI race, as well as billions of dollars of a return on its investment and a customer that will spend at least $250 billion on its services in the coming years.

Now, the company is looking elsewhere for help, and at the moment that means Anthropic. Don’t be surprised if there are others in the future as well.Will the Microsoft-Anthropic deal leave OpenAI out in the cold? – ComputerworldRead More