The bulk of US economic growth this year can be attributed to data centers. But what will all that infrastructure be worth once chips arrive that are a hundred times more energy-efficient than today’s models?
There are an awful lot of them now, those data centers—as our correspondent at Floating Junk Unlimited observed. First it was the logistics boxes that reshaped the global landscape; this year, it’s the unstoppable rise of data center construction. According to economists at ABN Amro, most of the economic growth in the United States this year was driven by data centers: the AI boom is devouring data, computing power, and with it enormous amounts of energy and cooling water. Infrastructure investments—especially in energy producers—have already been showing remarkably strong results for several consecutive quarters.
It reminds me of the three cable providers that, around the turn of the millennium, dug up our street, closed it, and dug it up again—each to lay their own cable beneath our sidewalk. In hindsight, that turned out to be a largely redundant investment, rarely recovered, and followed by widespread bankruptcies in the sector a few years later.
Will things turn out differently this time with data centers? Last Wednesday, Het Financieele Dagblad reported on a new chip invented by Euclyd that promises to be one hundred times more efficient than Nvidia’s. Production will take place in South Korea, but engineers seem convinced—as is Peter Wennink, the former CEO of ASML. Not exactly a lightweight endorsement.
Let me take a rough guess and say that roughly 1 trillion dollar has been poured into data centers so far. What would they be worth now, after Wednesday’s news? Technically, all those installations are, of course, still 100 percent operational, but from an economic and environmental standpoint, they may already be largely obsolete. With drastically lower electricity consumption and reduced need for cooling water (less power, less heat, less cooling), no one will want those guzzlers anymore. They will become partly junk—waste—depending on how easily those countless Nvidia chips can be replaced by Euclyd-type chips or others that are surely just around the corner. Qualcomm has already announced a new design, and the rest of the world certainly isn’t standing still either.
Did all this impress Nvidia’s investors? Not in the slightest. On Wednesday, the company hit a market cap milestone of 5 trillion euro. Perhaps they were quietly shocked by that number themselves, but investors remained euphoric in their FOMO-driven world. Yet those extremely expensive and wildly profitable Nvidia graphics chips may now be closer than ever to the theme of this column series: Floating Junk Unlimited—potentially the most costly waste ever produced.
Wouter Weijand worked in asset management from 1983 to 2025, over forty years as (lead) portfolio manager in bonds, equities, real estate, illiquid investments, and ultimately as CIO.