Microsoft replaces OpenAI, Anthropic models with in-house AI to cut costs

On July 7, Bloomberg reported a significant shift in Microsoft's AI strategy: the company has begun replacing models from OpenAI and Anthropic in select software products with its own in-house model, called MAI. The affected products include Outlook and Excel, two of Microsoft's most widely used applications. Microsoft's motivation is straightforward—reducing AI costs and decreasing reliance on external providers whose services have grown expensive. The timing aligns with earlier comments from Mustafa Suleyman, Microsoft's AI executive, who said in a June 4 report that the company was actively seeking alternatives to Anthropic because the provider had become prohibitively expensive. Microsoft's position in the AI infrastructure market supports this transition. Through Azure AI, its cloud platform, Microsoft provides the computing power and resources needed to train and run AI models. This means the company has built both the hardware foundation and the expertise to develop and deploy its own models. As Microsoft's internal models have improved in capability, the company now believes it can reduce its reliance on outside partners while lowering overall AI usage costs. The shift signals confidence that its proprietary technology can perform the necessary tasks—language understanding, content generation, and user assistance—that previously required third-party models in these productivity tools.

Microsoft's shift to in-house AI models reflects a maturing internal capability that challenges the company's previous heavy reliance on external partners like OpenAI and Anthropic. The company operates Azure AI, its cloud platform that provides the infrastructure to train and run AI models, positioning it to develop proprietary solutions. By moving models into Outlook and Excel—widely used productivity tools—Microsoft is testing its MAI model in high-volume, real-world environments where performance and cost directly affect profitability. The June statement from Mustafa Suleyman that Anthropic had become "extremely expensive" signals both the rising cost of third-party AI services and Microsoft's confidence that its internal models can meet the same quality bar at lower expense. This transition is strategic: it reduces the company's vulnerability to price increases from competitors while strengthening the stickiness of its own platform.