If you use Copilot at work, the model answering your questions is probably not the one you think it is.
Microsoft has begun quietly routing some Copilot requests in apps like Excel and Outlook from OpenAI and Anthropic models to its own, homegrown MAI models. The company rolled out MAI-Code-1-Flash inside Copilot Business and Enterprise last month, and at Build it launched seven new MAI models, including its first in-house reasoning model, MAI-Thinking-1. If you're a Copilot user, those routing changes happen under the hood and aren't surfaced in the app, even though the shift has been publicly reported.
The reason is money.
"Anthropic is extremely expensive and I think many people are urgently looking for alternatives," Microsoft AI CEO Mustafa Suleyman told Bloomberg in June. "We pay a lot of money to Anthropic. So our goal is to reduce and ultimately eliminate that cost."
The scale of what Microsoft is trying to cut is bigger than it sounds. Roughly $281 billion of Microsoft's commercial backlog is tied directly to OpenAI, which is close to half of the company's total remaining performance obligations. Every Copilot query that pings a GPT model is a payment Microsoft would rather not make. And analysts project Microsoft's quarterly capex could reach around $37.5 billion as AI infrastructure spending rises, much of it driven by GPU and data-center build-out. When you're that far in on infrastructure, running someone else's model on top of it starts to feel absurd.
So Microsoft is building alternatives across the stack:
- MAI-Thinking-1 is Microsoft's first in-house reasoning model, aimed at the same territory as Claude and GPT's frontier tiers.
- MAI-Code-1-Flash is now the default in Copilot Business and Enterprise, replacing the OpenAI models that were doing that work before.
- A tuned MAI model for Excel matches GPT-5.4 at up to 10 times lower cost, according to Microsoft's own evaluations.
The Excel number is the one that matters. Excel is where Copilot actually lives for a lot of people, and if Microsoft can hit parity there at a tenth of the cost, the case for keeping OpenAI in the loop gets harder to make internally. Suleyman also said the MAI models are being co-designed with Microsoft's own Maia 200 chips, which he claims is already producing a 1.4x efficiency boost. Owning the chips, the models, and the product is a play Microsoft has never really made in AI before.
Suleyman told The Verge that the pivotal moment was renegotiating the OpenAI contract, which freed Microsoft to train models at a much larger scale using its own IP. That's the strategic shift most people missed. The Copilot swap is the first visible product of it.
Still, this only works if the models are actually good enough, and that's where the story gets shakier. Independent benchmarks from BenchLM.ai put MAI-Thinking-1 well behind Claude Opus 4.5, especially on agentic tasks where it scores 46 versus Opus's 62.6. The Decoder put it plainly: Copilot customers may end up paying the same price for weaker AI so that Microsoft can lower its own costs. And Copilot adoption is still under 4.5% after three years, with only 1% of licensed users touching it weekly. Making the underlying model cheaper does not fix the product that most customers still aren't using.
The same week Microsoft was spending billions on embedded engineers to get Copilot into customer workflows, it was also quietly swapping out the models underneath. Both projects are pointing at the same problem, which is that Microsoft has bet the farm on AI and the economics are not working yet.
Microsoft is running the AWS playbook on its own suppliers. Build a homegrown version of the thing you resell, price it aggressively against the original, and let the switching happen quietly through defaults instead of announcements. OpenAI and Anthropic still get the headlines, but the customer they cannot afford to lose is now spending tens of billions to become their competitor. Suleyman says he wants Microsoft to be one of the top four labs in the world. What he actually needs is for the model in your Excel sidebar to be cheap enough that nobody notices when it changes, and good enough that nobody notices either.
