Satya Nadella has never been the kind of leader who waits for the future to happen. He has always paid special attention to the small signs that arise long before a big change in technology, and now again he seems to be sensing a big change. He told Microsoft’s top executives in a recent internal memo that they should not only reconsider how the company employs AI, but also how its whole business model should change to fit it. He spoke with the weight of experience and the strain of not knowing what would happen. He warned that some of Microsoft’s most successful businesses could go away if the corporation doesn’t change how it does things from the ground up.
Rolf Harms, an executive whose influence goes deep in Microsoft’s internal history, is at the center of this shift. Harms wrote the well-known 2010 white paper “Economics of the Cloud,” which many people at Microsoft still say changed the company’s way of thinking as it moved to cloud computing. Nadella has now made him an advisor on AI economics, which is basically asking him to do the same thing he did fifteen years ago: examine the things that everyone else is taking for granted.
Nadella stated in a note to top executives that Microsoft needs to “quickly rethink the new economics of AI across the company,” just as it did with the cloud. This made me think about how the cloud era has changed Microsoft in so many ways, including how it had to adjust its culture, business model, and do some difficult self-reflection. Nadella said that the AI era needs the same kind of change, but it needs to happen far more quickly.
He said that the company’s current change is “building a new AI factory and family of Copilots and agents that drive diffusion and usage across the full stack.” The pictures are startling because they make it look like Microsoft is not just adding AI to its current products, but also building a system where AI is the main engine. And this time, the economics are more complicated because corporations in the industry are worried that big investments in AI infrastructure won’t pay off in the long run. Earlier this year, Microsoft cut down on its AI spending, which made investors nervous. But it swiftly picked up the pace again with big new deals with OpenAI and Anthropic. Nadella’s email makes it apparent that being careful isn’t enough to define Microsoft’s approach right now, when the competition changes virtually every month.
Harms’ work is especially important since it made Microsoft had to have some tough talks. His study, “Economics of the Cloud,” questioned long-held beliefs within Microsoft about how the company marketed software, built infrastructure, and served business clients. At the time, some executives said he was “throwing bombshells into their org,” but his response—”they’re already there, I’m just helping you find them”—marked a turning point in Microsoft’s move away from packaged software and toward large-scale cloud services. That change is now largely seen as one of the main reasons Microsoft stayed competitive over the last ten years.
Nadella’s feeling of urgency now is more than just strategic analysis; it’s a constant fear that even big, successful firms might fail if they don’t change. He was honest about the future of Digital Equipment Corporation, which used to be a leader in minicomputers but went out of business after missing important changes in personal computing, at a town hall meeting in September. Nadella said that example “haunts” him and told employees that “some of the biggest businesses we’ve built might not be as relevant going forward.” The president of one of the world’s most powerful companies showed a rare moment of weakness, but it made the change he sees seem much more serious.
Harms will keep reporting to Scott Guthrie, who is in charge of Microsoft’s Cloud + AI division. He will also give advice to Nadella and the rest of the senior leadership team. He will assist Microsoft figure out how much it should charge for AI-driven solutions, how to distribute them, and how much infrastructure costs. In his annual letter to staff, Nadella gave another hint about how he sees the company’s future. He said that Microsoft is changing from being a “software factory” to a “intelligence engine,” a corporation that gives people and organizations the tools they need to make their own AI solutions. The small change in wording shows a bigger goal: Microsoft doesn’t just want to make things that people use; it wants to make things that people build with.
In another recent episode of the “Dwarkesh Podcast,” Nadella talked more about this change, saying that Microsoft is going from a “per user” price approach to a “per agent” strategy. He remarked, “Our business, which is now a business that makes tools for end users, will become more like an infrastructure business that helps agents do their jobs.” You shouldn’t simply think of the per-user business as per user; you should also think of it as per agent. This means that AI agents, which can make decisions, do tasks, and function on their own, will become economic units in Microsoft’s ecosystem. Microsoft sees a future when businesses pay for digital workers to work alongside their human colleagues, not just for people to use them.
