People seem to think that the “AI bubble bursting” somehow means “AI goes away.” Stop and think. There was a dot.com bubble. It burst. Some dot.coms died. Others flourished in the aftermath. You are more dependent on dot.coms today than you were before the bubble popped. The pundits who said the internet would not be profitable and would largely go away were completely mistaken.
So, why is that? The companies themselves were overvalued, not the technology or the market. Investors had assumed that those companies were nearly guaranteed future profits. That drove larger investments, which in turn drove larger valuations. Rinse/repeat until investors noticed the returns weren’t materializing as quickly as they expected. Valuations collapsed. The internet did not. The opportunities and the markets were still there. What was left was an enormous amount of infrastructure, software, intellectual property, engineering talent, and business experience. Much of that investment became a sunk cost.
This is one of the things bubbles often do. They finance the rapid build-out of infrastructure far beyond what can be justified by near-term demand. Investors lose money, but assets are left behind that can be used for decades. Railroads, fiber-optic networks, and internet infrastructure all had bubbles. They did not vanish because their investors overpaid for them. They became the foundation on which later companies built profitable businesses.
I expect the AI bubble to follow a similar pattern. Some AI companies will fail. Some valuations will collapse. Some investors will lose fortunes. But the GPUs, hyperscale datacenters, models, algorithms, software, research, and engineers will still exist. None of that disappears just because the stock market reprices the companies that built them. Apple, with its cash reserves, will probably be a beneficiary of the collapse by being able to cherry-pick technology and talent at fire-sale prices.
A bubble is a statement about asset prices. It is not necessarily a statement about the long-term value or viability of the underlying technology.
