Microsoft Spent Billions on OpenAI. This Month It Built Its Own Brain Instead.

Summary: Microsoft just shipped a reasoning model built from scratch and a new layer for developers to build agents on company data, reducing its dependence on OpenAI. The lesson for every business is not to train your own model. It is to own the one asset that off the shelf AI cannot copy.

Microsoft and OpenAI have been the defining partnership of the AI era. Microsoft poured billions into OpenAI, wired its products to OpenAI's models, and became the commercial face of the most famous AI lab in the world. So the most revealing AI story of the month is that Microsoft just spent serious effort to need OpenAI less.

This month Microsoft launched MAI-Thinking-1, a reasoning model trained from scratch and optimized for code, that the company says rivals the best models on hard benchmarks. The CNBC headline said the quiet part out loud: the models are designed to lessen reliance on OpenAI and lower costs for developers. Alongside the model, Microsoft is making its Work IQ APIs generally available, a layer that lets developers build agents on top of Microsoft 365 data, your email, calendar, meetings, files, and organizational context.

I write and speak about AI for business leaders, and I want to separate the two moves because they teach different lessons. Building its own model is about reducing dependence. Building the Work IQ layer is about owning proprietary data. Together they are a single strategy, and it is the most important strategic lesson of the week for any company, regardless of size.

Independence is the new moat

Microsoft did not build its own model because OpenAI's models are bad. They are excellent. Microsoft built its own because depending entirely on a single supplier for your core capability is a strategic risk, no matter how good that supplier is. This is the same lesson from the foundation model race, where Anthropic just passed OpenAI on valuation and revenue. The leaderboard moves. The supplier you depend on today may not be the best, or the cheapest, or even the most aligned with your interests tomorrow. If the most powerful company in software thinks single supplier dependence is too risky, the mid sized company betting everything on one vendor's API should think hard.

The asset a model cannot copy

The second move, Work IQ, points at something even more important. Microsoft is making it easy to build agents on top of Microsoft 365 data because that data is the thing competitors cannot replicate. Anyone can rent the same model. Nobody else has your company's emails, your calendars, your files, your accumulated organizational knowledge. The model is a commodity. The data is the moat.

In the Hive Structure, the bees are interchangeable. Any company can buy bees. What makes a hive valuable is the beekeeper's knowledge of where to point them, and the unique flowers only that hive can reach. Your proprietary data is the flowers. Your workflows and customer relationships are the field. Microsoft just built a product whose entire premise is that the data you already own, activated by AI, is worth more than the AI itself. That premise is true for your company too, whether or not you have noticed it.

What to do this week

Run two exercises. First, the dependence audit. Find every place your business relies on a single AI supplier with no alternative. For each, ask how hard it would be to switch. Anywhere the answer is very hard, you have a leverage problem Microsoft just spent billions to avoid, and you can address it with an integration layer for a tiny fraction of the cost. Second, the moat exercise. List the data and processes your company owns that a competitor could not recreate, then ask whether your AI efforts are actually built on those assets or just on the same public models everyone else uses.

The most powerful company in software just told you, through its actions rather than its marketing, exactly how to think about AI strategy. Do not depend on one supplier for your intelligence. Do build on the one asset no supplier can copy. The model is rented. The moat is yours. So which one is your AI strategy actually built on right now?

Sharon Gai is an AI transformation strategist, keynote speaker, and author of How to Do More with Less Using AI. She advises Fortune 500 companies on AI adoption and organizational redesign.