As the artificial intelligence race intensifies, technology giants are no longer focused solely on building smarter systems. The bigger challenge now is controlling the enormous costs that come with running advanced AI models at scale.
Microsoft appears determined to tackle that issue head-on, with the company now openly looking to reduce its reliance on Anthropic's Claude AI models.
In a recent interview with Bloomberg, Microsoft AI chief Mustafa Suleyman revealed that the company wants to eventually stop spending heavily on Claude altogether. "Anthropic is extremely expensive and I think many people are urgently looking for alternatives," Suleyman said.
The statement signals a major strategic shift inside Microsoft as it pushes to develop its own powerful AI systems capable of competing with leading models while operating at a lower cost. The company already spends significant amounts on AI services powering products like Copilot and supporting internal engineering teams. According to Suleyman, several departments within Microsoft are burning through millions of dollars in AI token costs, making efficiency a growing concern.
To address this, Microsoft recently introduced seven new AI models during its annual Build developer conference. One of the most talked-about releases was MAI-Thinking-1, a reasoning model the company claims performs at a level comparable to Anthropic's Claude Opus 4.6 on a widely used coding benchmark, but at a lower operational cost.
"We pay a lot of money to Anthropic, so our goal is to reduce and ultimately eliminate that cost," Suleyman said.
The rising expense of enterprise AI is becoming a serious issue across the tech industry. As companies integrate AI into coding, productivity, and business workflows, usage bills are climbing rapidly. Uber recently disclosed that it exhausted its entire 2026 AI coding budget within only a few months. Another company reportedly received a staggering $500 million invoice in just 20 days after employees used Claude without spending restrictions.
These incidents have triggered concerns among businesses deploying AI at scale. Many firms are now introducing usage limits and tighter controls after discovering that unrestricted AI access can quickly spiral into massive expenses.
Suleyman also claimed Microsoft has achieved significant efficiency gains while refining models for consulting firm McKinsey. According to him, Microsoft managed to outperform OpenAI's GPT 5-5 in some cases while delivering better cost efficiency. Reports further suggest the company has explored partnerships with Adobe involving its internally developed AI systems.
Microsoft's aggressive push against Anthropic also reflects changing dynamics in the AI market. Until late 2025, Microsoft's agreement with OpenAI limited its ability to independently pursue frontier AI development. A revised partnership now allows Microsoft to build competing AI models while still retaining access to OpenAI technology through 2032.
That arrangement gives Microsoft a unique advantage with OpenAI because it continues to benefit from discounted access to those models. Anthropic, however, does not offer the same long-term financial flexibility, making Claude a more expensive dependency for Microsoft.
At the same time, Anthropic's Claude models have become increasingly popular for enterprise coding applications, an area Microsoft considers strategically important. If Microsoft can offer similar performance through Azure and its own AI ecosystem at a lower price, it could become a compelling alternative for businesses seeking to reduce AI spending.
Still, Microsoft acknowledges the competition remains intense. Suleyman admitted that Anthropic has already moved ahead with newer and more advanced AI systems. While Microsoft believes it has made rapid progress, maintaining that momentum in a fast-changing AI industry may prove to be the company's toughest challenge yet.