Microsoft is reportedly moving more artificial intelligence work inside its own model ecosystem, reducing some reliance on OpenAI and Anthropic as the cost of powering AI features across productivity apps becomes harder for major tech companies to ignore.
Microsoft Turns to Its Own MAI Models
The shift centers on Microsoft’s MAI models, the company’s internal artificial intelligence systems.
TechCrunch reported that Microsoft has begun using its homemade MAI models to respond to a certain percentage of user prompts in Excel and Word.
The move does not mean Microsoft is cutting off outside AI partners completely.
Ground News said the change is incremental rather than a clean break, with OpenAI and Anthropic still handling most production traffic inside Copilot.
AI Costs Push Tech Firms to Get Thriftier
The decision reflects a broader industry problem: AI is expensive to run at scale. As AI costs continue to rise, companies are looking for ways to cut back.
Ground News reported that tech companies are acting significantly more thrifty after a brief period of “tokenmaxxing,” as the high cost of AI services pushes firms toward more affordable in-house solutions.
That pressure is especially important for Microsoft because AI is now embedded across everyday productivity tools. The more users ask Copilot to summarize documents, write emails, analyze spreadsheets or generate drafts, the more model usage becomes a real operating cost rather than a future investment story.
Build Conference Showed Microsoft’s Model Ambitions
Microsoft has also been preparing for this shift publicly. Microsoft announced seven new MAI models at its annual Build conference last month, including an agentic coder and a text-to-image generator. Microsoft unveiled seven MAI models at Build, including an agentic coder, and said one model tuned for McKinsey outperformed OpenAI models on cost efficiency by a factor of ten.
Those details show that Microsoft is not only looking for cheaper AI. It is trying to build a model portfolio that can route different tasks to different systems depending on cost, performance and data-control needs. That could matter for enterprise customers that care about where data is processed and how much AI features cost to operate at scale.
Reducing Dependence on OpenAI
The move also highlights Microsoft’s complicated relationship with OpenAI. Microsoft remains one of OpenAI’s most important partners, but relying too heavily on one external AI provider can create strategic risk.
CEO Satya Nadella reportedly fears Microsoft becoming “the next IBM” if it relies too heavily on a single partner, pushing the company to own more of its model technology rather than rent it. A 2025 contract renegotiation ended Microsoft’s exclusivity, freeing the company to build competing models while retaining a license to OpenAI’s technology through 2032.
Microsoft previously advertised that large parts of Office 365 were powered by models from both OpenAI and Anthropic. Now, the company appears to be shifting toward a more flexible model strategy where it can mix outside frontier models with cheaper internal systems.
A Wider AI Cost-Cutting Wave
Microsoft is not alone in trying to control AI spending. Other large companies, including Amazon, Uber, Meta and Accenture, have also reportedly moved to curb spending. The broader industry is seeking affordable in-house AI solutions as the cost of buying and providing AI services becomes a major concern.
The shift suggests that the next phase of enterprise AI may not be about using the biggest model for every task. It may be about using the right model for the right job at the lowest sustainable cost.
For Microsoft, relying more on MAI models could cut expenses, reduce dependence on outside providers and give the company more control over Copilot’s future. For the wider AI industry, it signals a practical turn: after the rush to add AI everywhere, companies are now asking how much those features actually cost to run.
