Microsoft is replacing OpenAI and Anthropic AI models in Excel, Outlook, and other applications with its own internally built models to cut costs. The shift marks a major strategy change as the software giant looks to reduce expensive licensing fees.

Microsoft AI Chief Executive Mustafa Suleyman told Bloomberg that “Anthropic is extremely expensive” and the company aims to “reduce and ultimately eliminate that cost.” The replacement applies to tens of thousands of AI prompts processed weekly in Microsoft’s core products.

Microsoft’s Cost Concerns

The company has grown concerned about the cumulative cost of relying on third-party AI providers. By moving to its own MAI (Microsoft AI) models, the company can improve margins and maintain more control over AI implementation across its product suite.

Microsoft is also training its sales team to position its AI offerings as more cost-effective and secure compared to OpenAI and Anthropic during customer conversations.

What This Means for the Industry

This move signals a broader shift in how large software companies view AI—not as a service to buy externally, but as core intellectual property to develop in-house. Microsoft’s confidence in its internal models suggests the company believes it has achieved sufficient capability parity.

The strategy could pressure OpenAI and Anthropic to lower prices or demonstrate superior performance to retain enterprise customers.

The AI market is entering a phase where cost efficiency and capability become table stakes for commercial deployment.

FYI (keeping you in the loop)

Why did Microsoft commit $2.5 billion to AI implementation?

Microsoft dedicated $2.5 billion and 6,000 employees to an AI implementation unit to ensure their AI strategy scales across products and customer deployments at lower cost than using external providers.