Microsoft has begun replacing AI models from OpenAI and Anthropic with its own in-house MAI models across several Copilot products, a strategic shift that is already live in Excel and Outlook and signals the company’s intent to reduce its dependence on third-party AI providers. The move raises a direct question for millions of subscribers: will you pay the same price for less capable AI so that Microsoft can cut its own costs?
Microsoft’s MAI Models Go Live in Excel, Outlook, and GitHub Copilot
The MAI models are now processing tens of thousands of requests per week in Excel and Outlook, both of which previously relied more heavily on models from OpenAI and Anthropic. The in-house models still handle only a small fraction of total Copilot requests, but Microsoft is steadily ramping up their deployment. MAI models are also available in GitHub Copilot, and a proprietary transcription model is expected to ship in Teams soon. The company’s head of AI, Mustafa Suleyman, openly described the logic in June: “We pay a lot of money to Anthropic—so our goal is to reduce and ultimately eliminate that cost.”
How MAI-Thinking 1 Compares Against OpenAI and Anthropic Models
At the Build conference, Microsoft unveiled seven new AI models, including MAI-Thinking 1, its first reasoning model. The company claimed it could match Sonnet 4.6 and Opus 4.6 in coding based on human evaluations. But the benchmarks released at the time told a different story. MAI-Thinking 1 trailed the competition from OpenAI and Anthropic by a wide margin and landed roughly on par with Deepseek V3.2. This performance gap matters because the same models are now being substituted into products that customers already pay for.
Why Microsoft’s Cost-Cutting Could Change What You Pay for Copilot
For Copilot and Office customers, the replacement of third-party models with in-house alternatives could mean paying the same subscription price for less capable AI. Microsoft CEO Satya Nadella has hinted that AI billing could shift further toward usage-based pricing instead of flat-rate subscriptions. One likely setup would make the cheaper MAI models the default tier, with third-party models from OpenAI or Anthropic available as premium add-ons at an extra cost. Under that model, Microsoft would pass its OpenAI and Anthropic expenses on to customers with a surcharge rather than absorbing them. This is the same Microsoft that recently argued vendor lock-in with OpenAI and Anthropic is a bad thing and positioned itself as a platform-neutral alternative.
The Data Training Contradiction
Microsoft has claimed that the MAI models are trained on clean, commercially licensed data, making them safe for enterprise use. According to the technical paper, however, Microsoft used the Common Crawl dataset—a collection of freely accessible web data whose use for AI training is not legally settled. Every other major AI company does the same thing, but Microsoft has portrayed its own training data as particularly clean. This distinction matters for businesses that need clear legal assurances about the data used to train the models embedded in their productivity tools.
What This Means for Copilot Subscribers Right Now
For individual professionals and enterprise customers, the practical takeaway is straightforward. If you are using Copilot in Excel, Outlook, or GitHub, you may already be interacting with Microsoft’s own MAI models rather than the more established alternatives from OpenAI and Anthropic. The performance difference is measurable, and if Microsoft moves toward usage-based pricing with premium tiers for third-party models, your cost could change depending on which model you need for your work. Monitor the model selection options in your Copilot settings and watch for any pricing announcements from Microsoft in the coming months. The era of a single, uniform AI tier inside Microsoft’s productivity stack is ending.
