They're already there. I'm just helping you find them.
In the long history of industries undone by their own success, Microsoft now finds itself at a familiar crossroads — not failing, but sensing the ground shift beneath its feet. Satya Nadella has summoned the architect of the company's cloud transformation to once again locate the hidden assumptions that could render its greatest strengths obsolete. The move from per-user to per-agent pricing is not merely a pricing adjustment; it is a philosophical reckoning with what kind of company Microsoft must become in a world where artificial intelligence, not human workers, may be the primary consumer of software. As with all genuine transformations, the danger is not ignorance of the future but attachment to a present that is still, for now, working.
- Nadella's internal memo landed as a quiet alarm — the company is thriving by every visible measure, yet he believes some of its largest revenue engines may not survive the AI transition without radical reinvention.
- The ghost of Digital Equipment Corporation haunts the executive suite, a cautionary tale of a dominant company that read the future too slowly and was erased by it.
- Rolf Harms, the economist who once dropped 'bombshells' into Microsoft's cloud-resistant divisions, has been recalled to find the next set of hidden contradictions buried in the company's current assumptions.
- The shift from per-user to per-agent pricing is already in motion — a structural change that redefines Microsoft's identity from an end-user tools company to an infrastructure provider for autonomous AI systems.
- The outcome remains genuinely uncertain: whether AI economics will prove as scalable and favorable as cloud economics did, and whether Microsoft can execute a second existential pivot, is a question the company is racing to answer before the market answers it first.
In early November, Satya Nadella circulated an internal memo to Microsoft's senior leadership carrying the weight of a warning: the company must rethink its entire business model for the AI era, and it must do so urgently. To lead that effort, he turned to Rolf Harms — the same executive whose 2010 white paper on cloud economics had forced Microsoft to confront uncomfortable truths about its future. The parallel was intentional. What cloud computing demanded of Microsoft fifteen years ago, artificial intelligence is demanding now.
Harms' original paper had been a rupture. It argued that customers would eventually overcome their skepticism about cloud security and availability, and that Microsoft had to restructure itself accordingly. The resistance inside the company was fierce — colleagues accused him of throwing bombshells into their organizations. His reply was characteristically calm: the bombshells were already there; he was simply helping people find them. Nadella is now asking him to find the next ones.
The urgency runs deeper than strategy. Nadella has said he is haunted by Digital Equipment Corporation, once a titan of the minicomputer era, undone by its failure to adapt. He has warned his team that some of Microsoft's most profitable businesses may not survive an AI-first world unchanged. In a July letter to employees, he named the strange paradox the company inhabits — growing and laying off simultaneously, thriving in an industry he described as having no franchise value.
The business model shift is already taking shape. Microsoft is moving from per-user pricing, the foundation of its empire, toward per-agent pricing — a recognition that AI systems are increasingly performing autonomous work alongside humans, changing the fundamental unit of value. Microsoft is becoming less an end-user tools company and more an infrastructure provider for agents that complete tasks and make decisions independently.
Harms will work under Scott Guthrie while advising Nadella and the broader executive team directly. His mandate is to help the company navigate what Nadella calls the new economics of AI. Whether this pivot will prove as successful as the cloud transformation remains unresolved — but the seriousness with which Nadella is asking the question suggests he believes the answer will define Microsoft's next chapter.
Satya Nadella has brought back an old architect. In early November, the Microsoft CEO circulated an internal memo to his top executives with a message that landed like a warning: the company needs to rethink its entire business model for artificial intelligence, and it needs to do it fast. To guide that reckoning, he turned to Rolf Harms, the same executive whose 2010 white paper on cloud economics had forced Microsoft to confront hard truths about its future fifteen years earlier. The parallel was deliberate. "This platform shift is all about building a new AI factory and family of Copilots and agents that drive diffusion and usage across the full stack," Nadella wrote. The message was clear: what the cloud did to Microsoft's business in 2010, artificial intelligence is doing now.
Harms' original white paper had been a watershed moment. It made the economic case for why customers would eventually abandon their skepticism about security and availability to embrace large-scale cloud services. At the time, people inside Microsoft resisted the findings so strongly that Harms later recalled they accused him of "throwing bombshells into their org." His response was disarming: "they're already there, I'm just helping you find them." Now, a decade and a half later, Nadella was asking him to find the bombs again—this time hidden in the assumptions underlying Microsoft's current business.
The urgency reflects something deeper than strategic repositioning. Nadella has become preoccupied with the fate of companies that failed to adapt. At a September town hall, he told employees he is "haunted" by Digital Equipment Corporation, which once dominated the minicomputer market before strategic missteps rendered it irrelevant. He has warned his team that some of Microsoft's biggest revenue engines may not survive the transition to an AI-first world without fundamental change. In a July letter to employees, he acknowledged the strange tension of the moment: the company was thriving by most objective measures while simultaneously conducting layoffs, what he called "the enigma of success in an industry that has no franchise value."
The business model shift is already underway, though most people outside Microsoft may not have noticed it yet. In recent interviews, Nadella has begun describing a fundamental reorientation in how Microsoft charges for its services. The company is moving away from per-user pricing—the model that built its empire—toward per-agent pricing. As artificial intelligence systems increasingly perform autonomous work alongside human employees, the unit of value changes. Microsoft is no longer primarily an end-user tools company. It is becoming an infrastructure provider that supports AI agents capable of completing tasks and making decisions independently. "Our business, which today is an end-user tools business, will become essentially an infrastructure business in support of agents doing work," Nadella explained. The shift is not theoretical. It is already reshaping how the company thinks about revenue, product architecture, and customer relationships.
Harms will report to Scott Guthrie, who oversees Microsoft's cloud and AI divisions, while working directly with Nadella and the executive team. His mandate is to help the company adapt to what Nadella calls "the new economics of AI"—a phrase that encompasses everything from infrastructure investment to platform technology to applications. The appointment signals that Nadella sees this moment as comparable in magnitude to the cloud transformation that saved Microsoft from irrelevance in the mobile era. Whether the company can execute another such pivot, and whether the economics of AI will prove as favorable as those of cloud computing, remains an open question. But the fact that Nadella is asking these questions now, with urgency, suggests he believes the answer will determine whether Microsoft thrives or fades.
Notable Quotes
This platform shift is all about building a new AI factory and family of Copilots and agents that drive diffusion and usage across the full stack.— Satya Nadella, in internal memo
Our business, which today is an end-user tools business, will become essentially an infrastructure business in support of agents doing work.— Satya Nadella, on the Dwarkesh Podcast
The Hearth Conversation Another angle on the story
Why bring back Harms specifically? Couldn't Nadella just hire any economist to study AI business models?
Harms isn't just an economist. He's the person who forced Microsoft to see what it didn't want to see about cloud computing. He has credibility inside the company because he was right before. That matters when you're asking people to abandon the business model that made them rich.
But cloud and AI are different technologies. Why assume the same playbook works?
It's not about the playbook being identical. It's about the pattern: a new platform emerges, the old business model becomes a liability, and the company has to choose between defending what works today or building what works tomorrow. Nadella is saying that choice is coming again, and he wants someone who understands the economics of that transition.
He mentioned being "haunted" by Digital Equipment Corporation. Is that just rhetoric, or does he genuinely fear Microsoft could disappear?
It's both. Rhetoric, yes, but grounded in real fear. DEC was the most dominant computer company of its era. It didn't disappear overnight. It died slowly because it couldn't let go of what made it successful. Nadella is trying to avoid that trap by acting before the trap closes.
The shift from per-user to per-agent pricing—that sounds like it could anger customers who are used to the old model.
It will. But Nadella seems to be betting that the alternative—staying locked in per-user pricing while AI agents do the work—is worse. If agents become the primary unit of value, then charging per user becomes economically nonsensical. Better to lead that change than be forced into it.
What does success look like for Harms in this role?
The same thing it looked like in 2010: helping Microsoft see what's already broken before the market forces it to see. If he does his job, people will complain that he's throwing bombshells into their organizations. That's how you know it's working.