📰 Key Highlights

Microsoft has recently started replacing some third-party models from OpenAI and Anthropic with its in-house MAI models in core Office software like Excel and Word to handle a certain percentage of user prompts. This strategic pivot, disclosed by Bloomberg, signals that Microsoft is actively reducing its dependence on external AI vendors and the related licensing costs. In fact, Microsoft had previously publicly promoted Office 365’s heavy reliance on OpenAI and Anthropic models as its technical foundation — a stance that has now quietly shifted.

On the in-house model front, Microsoft unveiled seven new MAI models at last month’s annual Build developer conference, spanning agentic coders and text-to-image tools, showing that Microsoft is simultaneously strengthening its own AI capability matrix.

Microsoft’s cuts aren’t an isolated case — they’re part of a broader cost-cutting wave sweeping the tech industry. After an initial “token maximization” spending frenzy earlier this year, major companies like Amazon, Uber, Meta, and Accenture have also reported trimming their AI spending. The enormous cost of AI services has become a flashpoint in Silicon Valley, with some companies even evaluating Chinese models for more competitive proxy pricing, despite the associated security concerns.

💬 JudyAI Lab Take

Microsoft slotting its in-house MAI models into Excel and Word and quietly swapping out OpenAI and Anthropic services — this isn’t just a technical choice. It’s a public signal from a platform player reclaiming control of its AI stack.

The most direct takeaway for AI builders from this case: your dependence on external APIs can be cut off at any time. Microsoft once loudly promoted Office 365 as being built on OpenAI and Anthropic, and has now quietly pivoted — the fundamental driver is licensing costs. And this kind of move isn’t isolated — Amazon, Uber, Meta, and Accenture are all pulling back AI spending, with some companies even evaluating Chinese models to drive down pricing, despite the security concerns. When “cutting costs” becomes Silicon Valley’s consensus, if downstream products only rely on “wrapping APIs,” once a vendor shifts strategy, whether their differentiation can hold up becomes a question you have to face head-on.

Now’s a good time to ask yourself: if the underlying model changes, does your product’s core value still hold? The answer determines where you should invest next.

📅 Original Article Info

🔗 Further Reading