Microsoft is reportedly changing how some AI features are powered inside Office 365, leaning more heavily on its own in-house models as the cost of AI services remains a pressure point across the technology industry.
The reported shift does not mean Microsoft has stopped using outside AI providers. Instead, it points to a more mixed strategy: keep using third-party models where needed, while routing some user prompts in Excel and Word to Microsoft’s own MAI models.
Microsoft is putting MAI models to work in Office
According to Bloomberg, Microsoft has begun using its homemade MAI models to answer a certain percentage of user prompts in two of its most widely used programs: Excel and Word.
That matters because Office 365 has previously been promoted as relying in large part on models from OpenAI and Anthropic. Microsoft still depends on those third-party systems, but the company is now reportedly adding more of its own technology into the product experience.
The change is best understood as an operational adjustment rather than a public product relaunch. Users may still see AI features inside familiar Office tools, but the model handling some of those requests may increasingly come from Microsoft’s own stack.
Cost pressure is shaping AI decisions
The timing fits a broader industry concern: AI is expensive to provide and expensive to buy. As AI costs continue to rise, companies are looking for ways to limit spending without stepping away from AI products entirely.
For Microsoft, relying more on MAI models could help reduce dependence on outside software for some workloads. It also gives the company more direct control over how certain AI agents and features are deployed across its own products.
TechCrunch reached Microsoft for comment, and the company said that it had "nothing further to share." That leaves the reported strategy as the clearest available signal: Microsoft is still using OpenAI and Anthropic models, but it is also building more of its own capacity.
The MAI push is broader than Word and Excel
Microsoft’s in-house AI work is not limited to Office prompts. Last month, at its annual Build conference, the company announced seven new MAI models.
Those models included an agentic coder and a text-to-image generator. That announcement shows Microsoft is trying to stand up more of its own AI agents across different kinds of tasks, from coding to image generation.
The reported use of MAI models in Excel and Word gives that strategy a practical outlet inside high-usage software. Rather than keeping MAI models only as research or platform announcements, Microsoft appears to be putting them into everyday productivity products where AI demand can be frequent and costly.
A wider cost-cutting pattern in tech
Microsoft’s reported move is part of a larger turn toward thrift among major technology companies. TechCrunch describes the recent news cycle as filled with stories about companies becoming significantly more careful with spending after a brief burst of “tokenmaxxing” earlier this year.
Other large companies have also reportedly moved to curb spending, including Amazon, Uber, Meta, and Accenture. The common thread is not that these companies are abandoning AI or large technology programs. It is that the cost side of AI is becoming harder to ignore.
The expense of delivering and purchasing AI services has become controversial in the industry. In some parts of Silicon Valley, the sticker shock has reportedly become severe enough that some companies are looking at Chinese models for more affordable agentic solutions, despite concerns over potential security issues.
What the shift signals
The clearest implication is that the AI market is entering a more cost-conscious phase. Large companies still want AI capabilities, but they are scrutinizing how those capabilities are delivered and what they cost at scale.
For Microsoft, the MAI strategy may help balance several priorities:
- Continue offering AI features inside Office 365 tools such as Excel and Word.
- Reduce reliance on OpenAI and Anthropic for some user prompts.
- Expand the role of Microsoft’s own AI agents and models.
- Respond to industry-wide pressure to control AI spending.
The shift also highlights a practical reality of AI deployment. Model choice is not only about capability; it is also about cost, availability, product fit, and control. As more companies move AI from demos into widely used software, those operational questions become central.
Microsoft’s reported use of MAI models in Office is therefore less about a single vendor change and more about the next stage of AI adoption. The technology is still advancing, but the business question is getting sharper: how can companies keep AI features useful while making the economics work?