Tech giant Microsoft has begun strategically replacing external artificial intelligence models from OpenAI and Anthropic with its own MAI (Microsoft AI) developments in key office applications — Excel and Outlook. This move is aimed at radically reducing the costs of expensive AI computations.
According to my data, tens of thousands of requests per week are now being processed directly within these applications using Microsoft's internal models. This marks a fundamental shift in the corporation's AI strategy, which has long relied on partner solutions with preferential access terms.
Why is Microsoft scaling back its collaboration with OpenAI and Anthropic?
The team led by AI division head Mustafa Suleyman is seeking to eliminate price dependency. Currently, Microsoft receives preferential access to computing power thanks to its long-standing partnership with OpenAI, but these discounts could be revoked in the future. To avoid finding itself in a vulnerable position, the company is accelerating the transition to its own models.
MAI's share of the total AI workload is still small, but the trend is clear. As early as June, Suleyman directly stated plans to cut costs on Anthropic and gradually transition to internal systems. "We pay Anthropic a lot of money — our task is to gradually reduce costs and completely eliminate unnecessary expenses," he emphasized.
This decision fits into Microsoft's overall cost-cutting drive. As early as May 2026, the company began a phased discontinuation of most internal Claude Code licenses. The transition to its own models is just part of a large-scale cost reduction.
Notably, despite record investments in AI, Microsoft is simultaneously cutting spending in other areas. As I previously noted, the corporation is laying off 2.1% of its workforce — 4,800 people, with the main blow falling on the Xbox gaming division, where about 3,200 specialists will lose their jobs.
Cryptalist Analysis: This is a classic example of vertical integration in the era of the AI race. Microsoft understands that long-term dependence on third-party providers is not only a financial burden but also a strategic risk. Creating its own MAI models allows not only to control costs but also lays the foundation for total monopolization of the corporate AI stack. For the market, this is a signal: the era of "free" partnerships is ending, and every giant will build its own AI castle.