I have watched hype cycles from the inside. I built a fintech startup, raised capital based on growth projections, and eventually sold it. I know what it looks like when investors confuse a rising category with a rising tide that lifts every boat. The AI market correction happening right now is a textbook example of exactly that mistake.

Investors collectively bet that almost every tech company would benefit equally from the AI boom. They were wrong. The result is a market wipeout measured in trillions.

Nvidia is down. Alphabet is down. Meta is down. Broadcom is down. Semiconductor stocks broadly are taking a hit. This is what happens when the market wakes up and asks the question it should have been asking from the beginning: who actually captures the value here?

Deutsche Bank analyst Jim Reid put it plainly: "Nobody truly knows who the long-term winners and losers of this extraordinary technology will be." That admission, from a major financial institution, is more important than the stock prices themselves. It means the consensus narrative — AI is the next internet and everyone wins — was always a projection, not an analysis.

Here is the structure of the actual AI economy as I understand it. Value concentrates at the infrastructure layer: the chip manufacturers, the hyperscale cloud providers, and the model developers at the frontier. Below that layer, most companies are consumers of AI capability, not producers of it. Being a consumer does not automatically make you more valuable. It might just make you more expensive to run.

The investors who drove AI stocks to stratospheric valuations were treating AI like an upgrade to every existing business model simultaneously. A healthcare company with AI is worth more. A retailer with AI is worth more. An enterprise software company with AI is worth more. The implicit logic is that AI features are pricing power, moat, and margin expansion all at once.

Sometimes that is true. Mostly it is not. Microsoft's Copilot integration is a real product with real revenue potential. A legacy retailer bolting on a chatbot is probably just a higher operating cost with a better press release.

What this correction signals is not that AI is a bust. It signals that differentiation is arriving. The market is beginning, belatedly, to ask which companies have genuine AI moats versus which ones are riding the category. Companies that have proprietary data advantages, deeply integrated AI workflows, or infrastructure positions will be fine. Companies that were labeled AI stocks because they mentioned AI on an earnings call are going to have a harder time.

The really interesting question is what happens to the hyperscalers — Microsoft, Amazon, Google — who have spent tens of billions on AI infrastructure. They bet on enormous enterprise demand materializing. If that demand comes but concentrates with fewer customers than expected, the infrastructure was overbuilt. We saw this story in the dot-com era.

I am not predicting an AI winter. The technology is real. The applications are real. Some companies are building durable businesses on top of it. But the correction is healthy. It is the market doing what it should have done earlier: asking hard questions about unit economics, defensible differentiation, and realistic timelines.

The technology is genuinely impressive. The question that should have been asked from day one is whether every company with an AI press release needed to be worth what investors were paying.