When Microsoft announced it would spend $80 billion on AI infrastructure in a single fiscal year, most people read it as a verdict. A company that had already decided where to place its chips. But listen to the extended interview Ben Thompson of Stratechery conducted with Satya Nadella around the same time, and a different picture emerges. Nadella kept returning — from multiple angles — to the question of what Microsoft is actually good at. Whatever the investment figure, he was still refining where Microsoft belongs in the AI ecosystem.
A CEO deploying hundreds of billions of dollars who is still working out his company's core competencies — I didn't read that as a sign of anxiety. I read it as a remarkably candid admission. And inside that candor, there's a thread that solo founders and independent PMs should pull on.
Between Partner and Rival
The topic Nadella handled most carefully in the interview was Microsoft's relationship with OpenAI. Microsoft has poured billions into OpenAI, and OpenAI's models run on Azure — a partnership on its face. But as OpenAI signs direct enterprise contracts at scale, the overlap with Microsoft's own Copilot and Microsoft 365 suite grows harder to ignore.
Nadella didn't dodge the tension. He gestured toward positioning Microsoft at the infrastructure and application layers rather than the model layer — but was measured about how stable that boundary actually is. Where Microsoft is currently placing its clearest bets is inference infrastructure: the more enterprises use AI, the more compute they need, and that demand flows back to Azure revenue. The working thesis is that Microsoft owns the floor on which software and AI agents run.
Where Nadella was noticeably less certain was the concept of an "agentic platform" — the stage where AI moves beyond assistant to active operator, executing work rather than supporting it. If that shift happens, the role of legacy software like Windows and Office changes fundamentally. Nadella called it the defining competitive terrain of the next decade, but on what it actually looks like: "still experimenting." The CEO of the company with more AI resources than almost anyone else on earth saying he doesn't yet know what the future platform looks like is a signal that this territory is genuinely open.
The Question That Arrives After You've Collected All the Tools
There's a tempting read of this interview that dismisses it as big-company strategy talk. Solo founders and independent PMs can pick up Notion AI, Perplexity, and Claude without any of this strategic deliberation, the argument goes — and when AI tools are this accessible, execution speed matters more than positioning. That's not wrong. The productivity shift from adopting a capable AI tool on day one is real.
But Nadella's concern is aimed at what happens after that fast-adoption phase plays out. When every competitor has the same tools, and every tool can handle the same tasks, the question "why am I the one who does this well?" changes weight. Microsoft has Azure, Office, and GitHub Copilot — and is still asking where it needs to stand on top of all of them.
When AI drafts content, organizes data, writes emails, and summarizes meetings as a matter of routine, the slice where a human adds something has to be defined more precisely and defended more actively. The same terrain opens up for solo PMs and independent creators.
Here's a pattern that shows up often in sales: someone is working hard, doing everything right — and deals keep not closing. Looking back, the problem usually wasn't effort. It was not knowing which specific thing they actually contributed. Some people's edge is relationship-building. Others interpret information better than anyone else. Others read timing. When AI starts handling a meaningful portion of those contributions, someone who never pinpointed their own edge will find that better tools don't translate into better results.
What to Figure Out Before You Stack Another AI Tool
Pull Nadella's interview into a solo-founder frame, and a set of concrete questions comes into focus.
Which parts of your work is AI already handling? As AI adoption expands, the line between tasks that can be processed quickly and tasks that require human judgment, trust, and experience gets sharper. Knowing where that line falls in your own work is the starting point.
From there: where do your results actually come from? If you're a content creator, you need to be able to distinguish whether your audience responds to your topic selection, your information architecture, your voice, or your publishing timing. Knowing which of those AI can substitute for — and which it can't yet touch — is what defines your actual contribution.
One thing Microsoft is watching in its OpenAI relationship is how deep dependency on a single partner narrows strategic flexibility. The same dynamic applies to solo operators: the moment an entire workflow is locked to one tool, that tool's policy changes or price increases can ripple across the whole business. Keeping substitutability in mind — regardless of scale — is sound preparation.
The question Nadella is still refining while spending $80 billion is, at its core, a foundational one. Building your tool stack first and understanding what you're genuinely good at first are two different sequences. When he says "still experimenting," he's signaling that the question stays open. For the solo PM, the available move is to arrive at your own answer before the field settles around you.
