$NBIS: A Great Business or a Great Story? I’ve spent a considerable amount of time thinking about $NBIS, particularly after reading some of the extraordinarily optimistic projections circulating on X. There’s no shortage of people explaining how many gigawatts the company might operate, how much revenue those gigawatts could generate, or why the stock is supposedly trading at some ridiculously low multiple of future sales. I understand the excitement, and frankly, I admire what management has accomplished in a relatively short period of time. But I think there’s a much more interesting conversation to be had, one that goes considerably deeper than revenue projections, capacity announcements, and the usual arguments about how enormous the AI opportunity will become. Let me make something clear from the beginning. I’m not suggesting $NBIS is a bad investment, that its ambitions are unrealistic, or that its shareholders won’t make substantial amounts of money. In fact, it’s entirely possible that the company becomes enormously successful and today’s valuation turns out to be attractive. What interests me is something much more fundamental: the difference between successfully building an extraordinary business and capturing an extraordinary share of the economic value that business helps create. There’s a tendency among investors to assume that the more important an industry becomes, the more valuable the companies operating within it must become. It sounds perfectly logical, and sometimes it turns out to be true. But economic history suggests something considerably more complicated, because the industries that transform civilization aren’t necessarily the industries that provide the greatest returns to their investors. Sometimes the enormous value created by an innovation flows primarily to customers rather than to the companies responsible for delivering it. And this brings me to the first question I would ask about $NBIS. If artificial intelligence becomes dramatically more powerful, computing capacity becomes increasingly abundant, and the cost of intelligence falls by 90% over the next decade, who actually captures the resulting economic value? Is it the infrastructure providers, the chip manufacturers, the businesses developing AI applications, or the millions of companies that suddenly become far more productive? I suspect the answer will be distributed very unevenly, and understanding that distribution might be considerably more important than forecasting how many gigawatts the industry eventually operates. Consider something fascinating about technological progress. When a product becomes dramatically cheaper, the amount consumed can increase enormously, sometimes by orders of magnitude. But an explosion in consumption doesn’t necessarily mean an explosion in profits for the companies supplying that product. In certain industries, the very process that makes the technology indispensable also makes it increasingly difficult for suppliers to maintain exceptional pricing power. Imagine computing demand increases tenfold over the next decade while the revenue earned per unit of useful computing falls by 80%. The industry would experience an extraordinary technological expansion, yet revenue per comparable unit of demand would have compressed dramatically. Total industry revenue could still double under those assumptions, but whether profits increased would depend on the costs of supplying that additional capacity. This is an illustration, not a forecast, but it highlights why demand growth and shareholder returns are two very different things. 1/👇