"Et par conséquent il s’ensuit que Dieu a choisi le meilleur des mondes possibles". (“And consequently it follows that God has chosen the best of all possible worlds.”)
— Gottfried Leibniz, Théodicée (1720)
“You met me at a very strange time in my life.”
― The Narrator, Fight Club (1999)
"It costs a lot to look this cheap."
— Dolly Parton
The glacial lake outburst of capital pouring into artificial intelligence infrastructure—tens of billions into GPUs, data centers, power grids, and software platforms—has people chanting “bubble.” There is, as someone remarked to me this week, a bubble in people saying "bubble".
Commentators—myself included—warn of froth, pointing to soaring valuations, unprecedented hardware orders, and the risk of stranded assets if demand falters, GPU rental margins collapse, Taiwan gets invaded, or maybe investors simply become pissy and there is a valuation gestalt flip.
It is a very strange time in economic life, but let's be less reactionary for a moment. Technology bubbles aren't necessarily bad, as economist Carlota Perez argued in her classic 2002 book. Viewed through her framework of technological revolutions, this moment can look less like a bubble in the narrow financial sense and more like a late-stage installation surge. That, in her framing, is the period when financial capital overshoots, infrastructure is bought and installed at great speed and well ahead of actual demand, and most of it goes bust. The wreckage left behind is real. But so is the scaffolding for later productivity.
Losses as the Logic of Progress
This is the unsettling rhythm of modern economic growth, one cited by economists from Smith to Schumpeter, to Perez. Every major economic revolution has reached a point where capital destruction became the mechanism, even if accidental, for building the future. It takes a lot of defunct assets to build a new market.
Some examples:
- In the 1840s, post-boom railway companies bankrupted investors across Britain. By the 1850s, the same bankrupt networks underpinned national integration, industrial transport, and military logistics.
- In the 1990s, the dot-com boom went parabolic, only to collapse spectacularly in 2000. Yet the networks, fiber, and server farms left behind formed the backbone of the 2000s online economy, enabling everyone from Amazon, to Netflix, to training AI models based on easy access to online data.
The pattern is brutal but familiar: capital overbuilds, much of it is destroyed, but society keeps the infrastructure.
The Secular Problem of Technology
So, if markets are efficient, why bubbles? If innovation is progress, why ruin? Because, Perez and others rightly argue, the stranded and abandoned assets from a capex mania can become the building blocks for what comes next, from unlit fiber, to currently unkitted AI data centers temporarily being used for laser tag.
This is a compelling argument. We never know if we've built enough until we've built too much. A technological revolution can only be messy, built on a speculative fervor ahead of cash flows. That will lead to waste and vast losses—and huge future gains from the defunct assets. And however it disrupts our lives, that is okay.
There are at least two problems with this view, however.