Microsoft Is Pulling Away from OpenAI
Bloomberg dropped a scoop on July 14: Microsoft has been quietly swapping OpenAI and Anthropic models for its own MAI series inside Excel and Outlook.
Tens of thousands of weekly AI prompt requests now run entirely on Microsoft-built models. The scale was never disclosed before.
Plain and simple — Microsoft is cutting costs.
Suleyman Said It Out Loud
We pay Anthropic a significant amount each year. Our goal is to gradually reduce and ultimately eliminate this unnecessary expenditure.
That is Mustafa Suleyman, head of Microsoft AI, speaking at Build in June. No hedging.
At Build, Microsoft released seven in-house models. One coding model matches Anthropic Opus 4.6 in performance at a fraction of the cost.
GitHub Copilot Switched Too
Not just the Office suite. GitHub Copilot now runs MAI models under the hood.
Suleyman also said a proprietary speech transcription model will land in Microsoft Teams in the coming months.
From code to email to meetings, Microsoft is pulling AI capabilities back in-house, one by one.
But GPT-5.6 Is Still the M365 Copilot Headliner
It is not that simple. The same week the MAI replacement news broke, OpenAI announced GPT-5.6 as the preferred model across Microsoft 365 Copilot — Word, Excel, PowerPoint, Chat, and Cowork.
Microsoft is swapping in homegrown models while putting GPT-5.6 front and center. The dynamic is delicate.
Translation: OpenAI powers the flagship experiences, MAI handles the cost-sensitive edges. Nobody is burning bridges, but Microsoft's self-sufficiency playbook is now public.
Why This Matters
Microsoft is OpenAI's largest backer, having poured in billions. But even the biggest investor does not want to work for its vendor forever.
MAI's rise signals one thing: AI models are shifting from scarce resource to infrastructure. And infrastructure has to be owned.
For small and mid-size businesses the takeaway is clear — do not bet everything on a single model vendor. Mixing models is the new normal.