A Silicon Supercycle? Or Just the Same Old Chip Cycle

A Silicon Supercycle? Or Just the Same Old Chip Cycle

Something unusual is happening with semiconductor stocks. These are the companies that make the chips inside phones, laptops, and now the massive data centers running AI. Many of these stocks have surged, and some have more than tripled in a matter of months. South Korea's stock market, which is heavily weighted toward chipmakers, doubled in about a year. For context, its first doubling took roughly 20 years. A move like that gets people talking, and a real debate has broken out among smart, well-informed investors about what it means. One camp believes the chip business has fundamentally shifted to a steadier, more predictable growth story. The other camp, which includes "The Big Short" investor Michael Burry, believes this is the same boom-and-bust cycle the industry has run for 50 years, simply playing out on a larger scale. Both sides have real evidence and neither is being unreasonable.

Semiconductors vs. Hyperscalers: Cumulative Return

Feb 17 – Jul 20, 2026: Semiconductors (SOX, light blue) vs. Hyperscalers Index (UBS, dark blue). Source: UBS. Past performance is not indicative of future results.

Why chips have been considered risky

To understand the debate, you have to know why chips, and memory chips in particular, have always been viewed as volatile investments. The pattern has been remarkably consistent. Demand rises, prices climb, companies rush to build new factories to capture the profits, and then, almost every time, they build too much. Supply overshoots demand, prices crash, profits evaporate, and the industry waits for the next upswing. It has repeated in 2000, 2008, and 2018. One number captures it well.

The bear case: we have seen this before

Skeptics argue that what looks and feels new in this story is really the old pattern at a larger scale. Data centers need enormous quantities of specialized memory to train and run these models, and demand has genuinely exploded. That much is not in dispute. But skeptics point out that the same "this time is different" conviction preceded every prior bust. The valuations bestowed on these companies are also striking. Semiconductor stocks are trading at price-to-sales and price-to-book multiples rarely seen outside the 2000 dot-com bubble. And the same companies racing to expand, pouring hundreds of billions into new capacity, are arguably planting the seeds of the next oversupply.

There is a deeper concern underneath the rally as well. While the large technology companies buying these AI chips are spending staggering sums, many are not yet showing profits from AI that come close to justifying the outlay. If those customers do not eventually earn a return on what they are building, skeptics argue, the spending slows or reverses, and chipmakers are left holding factories built for demand that never fully arrives.

The bull case: something has genuinely changed

The other side is not supported simply by optimism from people who own the stocks. It rests on a real change in how these companies sell. In the past, memory makers sold mostly into an open spot market, where prices swung week to week based on what buyers were willing to pay. Today, some of the largest players have locked in years of business via binding contracts, in which the customer commits to buy a set amount at a set price whether they end up needing it or not. One major memory maker has already collected over $20 billion in upfront payments this way, with near-term production entirely spoken for. Proponents argue this is a fundamentally different arrangement than the old model, because revenue is locked in ahead of time rather than left to the mercy of the spot market.

The structural argument goes further. Building a new chip factory takes years, not months, and the new plants under construction now will not be running until 2027 at the earliest. In the meantime, more of the existing capacity is being redirected toward the advanced memory that AI requires, which leaves less available for the ordinary memory used in regular PCs and Apple devices. That supply squeeze is pushing prices higher across the board, and it cannot be resolved by simply deciding to make more chips. There is a genuine multi-year gap between what everyone wants and what can physically be produced. Some industry executives are calling it the largest supply-demand imbalance the memory business has seen in 25 years, and framing it as a lasting shift rather than another temporary spike.

So who is right?

Both sides are looking at the same facts and reaching different, reasonable conclusions. The skeptics are right that the industry has a consistent history of overbuilding after a boom, and that today's enthusiasm echoes, almost word for word, the enthusiasm that preceded past crashes. The optimists are right that something concrete really has changed, with contracts and physical bottlenecks that did not exist in the same way before.

Our view is that both are true, and not in the tidy sense that one side simply gets proven right a few years from now. They are true at the same time, in the same market. Some firms will overbuild, chase the demand curve, and eat the classic bust when supply catches up. Others have used contracts and disciplined capacity planning to lock in demand and position for the structural shift, and they will likely sidestep much of the damage. The cycle is not being cancelled but distributed unevenly across the companies competing in it.

That distinction matters, because the last few years trained investors to treat this space as a single trade. Buy the sector, own the theme, and let a rising tide lift every boat. We think those days are ending. When the tide no longer carries everyone, the gap between the companies that compound through the cycle and the ones that get caught overextended becomes the entire story. A sector-wide bet that worked when everything went up is a very different proposition when the winners and losers start to diverge.

Which is a long way of saying the work is back. Careful analysis of who holds real pricing power, who has genuinely locked in demand, and who is simply spending to keep up is as important as it has ever been. The names in this space will not move together the way they have, and telling the durable businesses apart from the ones running on hope is precisely the job in front of us.