The narrowest boom in history

Ten thousand people. That's roughly the size of a small English market town, and it's TechCrunch's estimate of how many AI insiders have hit extraordinary wealth in this cycle. Previous tech booms spread their winnings across far more people over longer timelines. This cycle is producing fewer winners, faster, at higher individual stakes.

The concentration problem

Every technology boom creates wealth inequality. What makes this one structurally different is the ratio of capital deployed to people needed. Cerebras burned through $8 million a month and almost died early on solving a chip design problem with extreme technical and financial risk. The reward: a $60 billion valuation. The team that got there was tiny relative to the capital at risk. That ratio is the story.

This is the defining feature of AI economics. The infrastructure is enormously expensive, but the organisations capturing that value employ remarkably few people. When Bill Ackman built a Microsoft stake during the Q1 sell-off, his thesis was about AI and cloud growth compounding inside an enterprise platform. He wasn't betting on Microsoft hiring more people. He was betting on the same workforce producing exponentially more value. The bet is on leverage, not labour.

The parallel that keeps coming to mind is the oil industry circa 1870. Standard Oil's dominance wasn't built on employing the most people; it was built on controlling infrastructure that everyone else depended on. The wealth concentrated in a handful of operators while the broader economy restructured around their output. We're watching something similar, except the timeline is compressed from decades to quarters.

What's uncomfortable for product builders is where this leaves the middle. The haves-and-have-nots divide in AI isn't between tech and non-tech anymore. It's between the few thousand people training frontier models and building inference chips, and everyone else, including most of the tech industry. The skills gap has become a wealth gap.

Cerebras's 10% stock drop on its first full trading day after the blockbuster IPO is instructive here. The market priced in the hype on day one and started pricing in the risk on day two. A $60 billion valuation for a chip startup is a bet that concentration works, that you only need a few massive buyers if those buyers are spending at civilisational scale.

I think the real question for anyone building products right now isn't whether AI will transform their industry. It's whether the transformation creates opportunities for thousands of companies or funnels value to a dozen. Ackman's bet on Microsoft's AI and cloud growth is essentially a bet on the latter: that the platform incumbents will capture most of the value. If he's right, the strategic question for everyone outside that inner ring isn't "how do I compete?" but "how do I build something valuable in the slipstream of platforms I'll never displace?"

The narrowest boom in history might also be the shortest window to find out.

---