When a good idea meets borrowed money

Artificial intelligence is real, and the money being spent on it is enormous. Datacentres are being built faster than global supply chains can keep up, and shortages have appeared in memory chips, networking gear, power equipment and more. Suppliers with spare capacity have been able to name their price, and their profits have been spectacular.

The distinction that is easy to miss: those profits come from scarcity as much as from demand.

Australians have watched this film before. In a mining boom, the price of iron ore runs far above what it costs to get the stuff out of the ground, and for a few years the companies that own the mines earn extraordinary returns. They have not become better businesses; they own something scarce at the moment everyone wants it. The profits then attract capital, new projects are approved, and because mines take years to build, the boom lasts long enough to start feeling permanent. It never is.

Memory chips work the same way. Memory is commodity-like, sold by the gigabyte, and a new fabrication plant takes years and billions of dollars to bring online. So the question is not whether the AI boom is real. It is what you paid for it, and what you quietly assumed about how long the shortage lasts.

Asian markets have just delivered a live demonstration. South Korea’s sharemarket roughly doubled over the first half of 2026 on the strength of its memory chip makers, gave back close to a third of that in a matter of weeks, then rebounded violently at the end of July. Nothing about the underlying technology changed.

What changed was leverage. Borrowed money had been pouring into a narrow group of shares, through margin lending, leveraged funds tied to single stocks and a wave of new trading accounts. When prices fell, lenders demanded repayment, forcing sales, which pushed prices lower still. Over a million Korean brokerage accounts received margin calls. Globally, one of the largest AI-focused funds in the world was compelled to liquidate its entire portfolio of listed shares within days. Leverage does not change what a business is worth. It changes who can afford to wait, and forced sellers do not get to choose their timing.

The quieter consequence is where all that money came from. Capital has been pulled out of the rest of Asia to chase one theme, and markets such as the Philippines now trade at valuations not seen since the global financial crisis — banks, property developers and consumer businesses priced for neglect rather than for anything going wrong in how they are performing.

None of this argues against artificial intelligence, likely one of the defining economic forces of the coming decades. But it argues for keeping two questions apart. Is this a great technology? and is this a good investment at this price? are not the same question — and in the excitement they are easily confused. The second one is usually answered somewhere the crowd isn’t looking.