Microsoft has significantly expanded the use of its own artificial intelligence models, known as MAI, to handle tens of thousands of prompts generated weekly in its Office applications Excel and Outlook. This move marks a clear shift away from reliance on external AI providers like OpenAI and Anthropic, as reported by Bloomberg citing unnamed sources.
The company’s strategy to substitute its own AI models for third-party technology is driven by the need to lower AI-related expenses and gain greater control over its AI infrastructure. Earlier statements from Microsoft AI’s CEO Mustafa Suleyman revealed plans to reduce spending on Anthropic-supplied models and to deploy Microsoft’s internal AI capabilities more broadly across its product portfolio, including transcriptions for the Teams videoconferencing app.
Reports from earlier in 2025 indicated that Microsoft had been testing alternative AI models capable of competing with OpenAI’s solutions, with training efforts already producing models that matched OpenAI’s performance on key benchmarks. Microsoft had maintained a multi-source approach to AI integration, employing a combination of OpenAI, Microsoft’s own models, and open-source technologies across its products.
This strategic pivot aligns with the company’s previous increased investment in Anthropic’s AI, which brought annual costs for those services to roughly half a billion dollars. Despite the rising expenditure on AI and cloud capabilities fueling growth, Microsoft has faced investor concerns about elevated capital spending, balancing innovation with cost management challenges.