Microsoft replaces OpenAI with its own AI in apps
▼ Summary
– Microsoft has begun replacing OpenAI and Anthropic models with its own MAI models in some product features where cost or data residency is advantageous, according to Bloomberg.
– The change is incremental; OpenAI and Anthropic still handle most Copilot traffic, while MAI models are used where they are more cost-effective.
– Microsoft unveiled seven MAI models at Build, including reasoning, image, voice, and transcription systems, and is testing them in Teams, Copilot, Bing, and PowerPoint.
– The shift is enabled by a 2025 renegotiation that ended Microsoft’s exclusivity with OpenAI, allowing it to build competing models while keeping a license to OpenAI’s technology.
– Microsoft’s strategy is to reduce reliance on single partners, as CEO Satya Nadella feared becoming “the next IBM,” and to improve margins by using cheaper in-house models on its own Azure infrastructure.
Microsoft has quietly started swapping out OpenAI and Anthropic models for its own in-house AI in select product features, according to Bloomberg. The move routes specific tasks to Microsoft’s MAI models when factors like cost efficiency or data residency make them the better choice.
This shift is gradual rather than a sudden split. For now, OpenAI and Anthropic continue to handle the bulk of production traffic within Copilot, with MAI models stepping in only where their economics prove more favorable.
At its Build conference, Microsoft unveiled seven MAI models, including its first reasoning model alongside systems for image, voice, and transcription. The company has been testing MAI-Transcribe-1 across Teams and Copilot, while rolling MAI-Image-2 into Bing and PowerPoint.
The core advantage is efficiency. By running these models on its own Azure infrastructure, Microsoft avoids paying third-party fees. One MAI model, tailored for consulting firm McKinsey, reportedly outperformed OpenAI’s GPT-5.5 on cost efficiency by a factor of ten.
This doesn’t mean a full divorce is imminent. Microsoft launched the MAI line as a direct rival to OpenAI while still keeping its partners’ models available for tasks where they excel.
For years, Microsoft’s original contract with OpenAI barred it from independently pursuing frontier AI. That changed in 2025 when the two companies renegotiated, ending Microsoft’s exclusivity and freeing it to build competing models while retaining a license to OpenAI’s technology through 2032.
The deal worked both ways, allowing OpenAI to sell through rivals like AWS. Microsoft’s response has been a three-way hedge: holding a large OpenAI stake, embedding Anthropic’s Claude in Copilot, and now shipping its own MAI models.
The strategic reasoning echoes what CEO Satya Nadella has hinted at directly. He reportedly feared Microsoft becoming “the next IBM” if it leaned too heavily on a single partner. Owning the model, not just renting it, is how the company avoids that fate.
Microsoft doesn’t need MAI to top every leaderboard. Its reach into hundreds of millions of Office and Teams seats does the heavy lifting. Shifting even a fraction of that traffic to first-party models moves real money.
That reach comes with a persistent challenge: the question of paying customers remains unresolved. Cheaper in-house models help the margins, whether or not users pay up.
The cost of outside AI isn’t trivial, and every feature Microsoft brings in-house trims that expense. The partnership that once defined the company’s AI story is being quietly rebalanced, one swapped-in model at a time.
(Source: The Next Web)
