Microsoft expands in-house AI models to reduce reliance on OpenAI, Anthropic

Microsoft is reportedly accelerating the use of its internally developed artificial intelligence models across its software ecosystem as the company works to reduce its dependence on external AI providers and lower the long-term cost of operating AI-powered services.According to a Bloomberg report, Microsoft has begun using its proprietary "MAI" models to handle a portion of AI requests generated through applications such as Excel and Outlook. While these workloads represent only a small share of the company's overall AI operations, the move signals Microsoft's growing confidence in the performance of its own models.Until recently, many of these AI-powered capabilities relied on foundation models from OpenAI and Anthropic. The reported shift reflects Microsoft's broader strategy to diversify its AI infrastructure as demand for generative AI services continues to rise.The transition comes as the company seeks to manage the high computing costs associated with AI. Every AI-generated response requires significant processing resources, making inference costs a critical factor for companies deploying AI assistants at scale. Although Microsoft has benefited from favourable commercial terms through its long-standing partnership with OpenAI, the company is reportedly preparing for a future in which those economics may change.Also read: SpaceXAI, Cursor set to unveil first jointly built AI model amid enterprise AI pushMicrosoft AI Chief Executive Mustafa Suleyman had earlier indicated that reducing dependence on external AI providers is a strategic priority. During the company's Build developer conference in June, Microsoft introduced seven new in-house AI models designed for a range of enterprise and developer applications.The company said one of its internally developed coding models delivers performance comparable to an earlier-generation coding model from Anthropic while operating at a lower cost, highlighting Microsoft's efforts to improve efficiency without compromising capability.Microsoft is also broadening the deployment of its proprietary AI models beyond productivity software. The report said the technology is already being integrated into GitHub Copilot, while an in-house speech transcription model is expected to power features in Microsoft Teams and other products in the coming months.