Microsoft Bets on In-House AI to Cut OpenAI and Anthropic Costs

Microsoft is developing its own proprietary AI models (MAI) to reduce reliance on OpenAI and Anthropic, routing tens of thousands of prompts weekly through its in-house technology. The company unveiled seven MAI models in June 2026, including MAI-Thinking-1, which matches Anthropic's Claude Opus 4.6 on coding tasks. This strategic move aims to convert expensive rented AI costs into owned infrastructure, improving ROI on Microsoft's AI spending while providing leverage in future negotiations with third-party AI providers.