Microsoft wants Anthropic out and does not want to spend heavily on Claude

Microsoft is working to reduce its reliance on Anthropic's Claude AI models, with AI chief Mustafa Suleyman calling them too expensive. The company believes its new in-house AI models can deliver similar capabilities at a lower cost.

In Short

• Microsoft wants to eventually eliminate spending on Anthropic's Claude
• The new MAI-Thinking-1 model is claimed to rival Claude Opus 4.6
• Anthropic has emerged as Microsoft's biggest external AI cost concern

The AI race is no longer just about building smarter models. It is increasingly becoming a battle over costs, and Microsoft appears ready to take on one of the biggest names in the industry for that reason. While many people still see OpenAI as Microsoft's biggest AI partner and rival, the company's AI chief Mustafa Suleyman has pointed to a different competitor that is currently on his radar - Anthropic. In a recent interview with Bloomberg, Suleyman stated that Microsoft wants to reduce its dependence on Anthropic's Claude models and eventually stop paying for them altogether.

"Anthropic is extremely expensive and I think many people are urgently looking for alternatives," Suleyman said.

His comments come at a time when businesses are discovering that running advanced AI systems at scale can quickly become a costly affair. Microsoft itself spends heavily on AI services across products such as Copilot and within its engineering teams. According to Suleyman, many teams inside the company are spending millions of dollars on AI tokens, making cost reduction a top priority. What Microsoft is trying to do is build powerful AI models internally, so that it doesn't need to spend large amounts of money on external providers like Anthropic.

Microsoft says Claude is becoming too expensive for enterprises

Microsoft recently unveiled seven new AI models during its annual Build developer conference. Among them is MAI-Thinking-1, a reasoning model that the company claims can deliver performance comparable to Anthropic's Claude Opus 4.6 on a popular coding benchmark while costing less to run.

"We pay a lot of money to Anthropic, so our goal is to reduce and ultimately eliminate that cost," Suleyman said.

For Microsoft, this is not merely about matching competitors. The company sees lower-cost AI as a necessity for long-term sustainability. As enterprises deploy AI across thousands of employees, the bills can quickly spiral. A good example of rising costs is what happened with Uber and another company. Uber recently revealed that it lost its entire 2026 AI coding budget in just a few months. In fact, a company recently went viral for receiving a $500 million bill for just 20 days after they forgot to put usage limits on Claude for employees.

The issue is already becoming visible across the industry. Companies are beginning to put limits on AI usage after costs exceeded initial expectations. With AI tools now deeply integrated into software development and workplace productivity, controlling spending has become as important as improving performance.

Suleyman also claimed that Microsoft achieved major efficiency gains while refining its models for consulting giant McKinsey. According to him, the company managed to outperform OpenAI's GPT 5-5 while delivering significantly better cost efficiency. Reports also suggest Microsoft has held discussions with Adobe about using its in-house AI models.

Why Anthropic and not OpenAI?

The timing of Microsoft's push is not accidental. Until late 2025, Microsoft's partnership with OpenAI restricted the company from independently pursuing frontier AI development. A revised agreement changed that situation, allowing Microsoft to develop competing models while still retaining access to OpenAI technologies through 2032. That arrangement gives Microsoft a major advantage when it comes to OpenAI. The company continues to enjoy discounted access to OpenAI models and does not face the same financial pressure there. Anthropic, however, is a different story.

Microsoft does not have a similar long-term arrangement with Anthropic, making Claude a more expensive dependency. From a business perspective, replacing Anthropic with Microsoft's own models could save substantial amounts of money while strengthening the company's AI ecosystem.

There is also a competitive angle. Anthropic's Claude family has become a popular choice for enterprise coding applications, an area Microsoft views as strategically important. If Microsoft can offer similar capabilities through Azure and its own AI products at a lower cost, many customers may find less reason to pay separately for Claude.

Still, Microsoft acknowledges that the challenge is far from over. Suleyman admitted that Anthropic has already launched newer and more advanced models since Claude Opus 4.6. While he believes Microsoft has made rapid progress in a short period, staying competitive in an industry where breakthroughs happen every few months will be the bigger test.

- Ends