Home / Technology / Microsoft Bets Big on In-House AI
Microsoft Bets Big on In-House AI
15 Jul
Summary
- Microsoft is routing Excel and Outlook prompts to its own MAI models.
- MAI aims to reduce Microsoft's significant costs paid to OpenAI and Anthropic.
- This strategy diversifies Microsoft's AI dependencies and strengthens its competitive position.
Microsoft is implementing a strategy to decrease its dependence on external frontier AI models. In June 2026, the tech giant began routing select prompts in applications such as Excel and Outlook to its own in-house AI models, known as Microsoft AI (MAI).
This initiative is driven by the substantial costs associated with using third-party AI services like those from OpenAI and Anthropic. Microsoft aims to convert these rental costs into owned infrastructure, thereby boosting profit margins. The company unveiled seven MAI models in June 2026, including MAI-Thinking-1, which reportedly matches Anthropic's Claude Opus 4.6 on coding tasks.
By developing and utilizing its own AI models, Microsoft is building a more robust competitive moat. This approach provides flexibility, hedging against potential future cost increases from partners and insulating the company from a singular point of AI dependency. The strategy ensures Microsoft maintains leverage in negotiations and avoids becoming overly reliant on other tech firms.
While this shift is incremental, with most Copilot workloads still relying on outside models, it signals a clear direction. Concerns remain about the quality comparison between MAI and current leading models. This move also presents a warning to AI labs like OpenAI and Anthropic, as their key distribution partner becomes a competitor.
