US semiconductor stocks are now the market’s real AI bet

US semiconductor stocks now account for almost a fifth of the total US stock market. That is not a normal sector rotation. It is a structural bet by investors that the next phase of economic growth will be built on chips.

The obvious comparison is the dot-com bubble. In 2000, semiconductor stocks also became one of the great proxy trades for the future. The internet was real. The demand was real. The mistake was not believing in the technology. The mistake was assuming that every dollar of excitement would become a dollar of durable profit.

That distinction matters today. Artificial intelligence is not a fantasy. The demand for compute is visible in data centre spending, cloud infrastructure, memory shortages, networking demand and the extraordinary growth of companies such as Nvidia and Broadcom. The Semiconductor Industry Association says global chip sales are on track to reach $1 trillion in 2026, with Q1 sales significantly above Q4 2025. Deloitte has also estimated that generative AI chips could approach $500 billion in revenue this year, roughly half of global chip sales.

So the market is not inventing the story. It is repricing the foundations of the digital economy. Chips are no longer just components inside devices. They are becoming strategic infrastructure: the picks and shovels of AI, cloud computing, robotics, defence, autonomous systems and scientific research. That is why this rally is more serious than the usual “tech bubble” dismissal suggests. The revenues are there. The customers are real. The spending is coming from some of the richest companies in the world. AI infrastructure is not being financed by retail speculation alone. It is being financed by Microsoft, Amazon, Alphabet, Meta and the rest of the hyperscaler economy.

But that does not make the trade safe

When one industry grows to nearly 20% of the US equity market, it stops being a sector story and becomes a market risk. Investors are no longer just betting on chip companies. They are betting that AI capital spending will continue, that margins will hold, that supply chains will remain open, that Taiwan risk will not become unmanageable and that customers will eventually earn a return on the billions they are pouring into compute.

That is a lot of assumptions for one trade to carry

The biggest risk is not that AI fails. The bigger risk is that AI works, but the economics become less generous than investors expect. Competition can rise. Pricing power can fall. Custom chips can reduce dependence on the current leaders. Export controls can limit sales. Energy constraints can slow data centre buildout. And at some point, the buyers of AI infrastructure will need to show that this spending produces enough revenue, productivity or cost savings to justify the scale of the boom.

This is where the comparison with 2000 becomes useful. The internet did change the world. But the stock market still paid too much, too early, for too many companies. A transformative technology can be completely real and still produce a valuation problem.

That is the story this chart tells. Semiconductors have become the market’s cleanest expression of the AI boom. They are also where much of the risk is now concentrated.

The US stock market has always had dominant sectors. Railways, oil, banks, telecoms and software have all had their moment. The difference now is that semiconductors sit at the intersection of corporate spending, national security and geopolitics. They are not just another growth sector. They are the industrial base of the AI age.

That makes the rally understandable. It also makes it dangerous.

If the AI buildout delivers, semiconductor stocks may deserve a permanently larger place in the market. But at 19.3% of US equities, investors are not waiting for proof. They have already priced semiconductors as the new centre of gravity.

That is the bet. Not that chips matter. They clearly do. The bet is that they matter enough to justify reshaping the entire US stock market around them.