The Semiconductor Rally Is Going Strong, but History Suggests the Cycle Still Matters

It’s hard not to feel just a hint of FOMO (the fear of missing out) when it comes to the semiconductor plays after their latest leg higher. It seems like the group is becoming unstoppable, especially as investors, including the smart money, pile into the hardware names that seem like easier and more obvious winners from the AI revolution as it heads into its next stages (agentics and robotics).

At the end of the day, the great Jensen Huang, the top boss over at GPU titan Nvidia (NASDAQ:NVDA | NVDA Price Prediction), put it best: modern AI data centers are “AI factories” that turn electricity into tokens. As the next wave of AI innovations levels up the token appetite, it feels like the chip side of things will continue to be a bottleneck where much of the earnings spoils will be.

The semis continue to be the hot place to be

Why take a chance on the potential AI plays in the application layer when it’s harder to tell what wins and what loses when you can just play the token side of things while getting earnings in the present and near future rather than having to wait for some unknown period of time before the ridiculously hefty CapEx starts showing signs of return?

Combined with the recent market pessimism surrounding software, it’s clear that greater certainty (in the semis) is a far better place to be. Why not bet on the incredibly profitable chokepoints of the AI revolution, then pick and choose names that are being forced to play defense against some AI-native rivals that are as scary as can be?

Any way you look at it, the semi rally could go either way over the near term, as it appears the recent spike higher has mostly baked in the token surge to come. But the big question, in my opinion, is whether the cyclicality of the semis still applies in the AI age. Of course, the argument for a long-lived structural surge makes sense.

Are things really different this time?

We’re in the midst of an AI revolution, and token consumption could skyrocket at a rate that’s difficult to comprehend. If we’re talking about 24/7 agents running all over the place, we’re going to need a heck of a lot more tokens. And that means the semis still have a ton of work to do.

A parabolic spike in the demand for tokens and Jevons paradox in play as the latest chips are worlds faster and more efficient than the last, may very well allow for more gains to be had within the overheated semi space. Is this time truly different? And is the semi landscape enjoying a structured multi-year ascent that only a transformative revolution could pave the way for? Or is this just a large cyclical boom that will be followed by a bust?

Given Michael Burry’s bearish bets against the semiconductors, it feels like a cyclical bust, and the possibility of a bubble cannot be ruled out.

For now, demand is off the charts, and it could take some time before supply catches up. If the next wave of AI tech keeps increasing token consumption, perhaps supply could be playing catch-up for a whole lot longer. It’s really hard to tell if things are different this time around, and the history should be discounted.

Past cycles have really hurt those who were late to the party. At the same time, though, there is perhaps no good comparable that lies in the past. The AI revolution is certainly magnitudes bigger than the internet boom. And with firms focused on monetization, perhaps a profitability surge could stomp out the bubble fears.

The bottom line

While I respect Michael Burry, I do think that odds favor more of an AI supercycle or even a structural surge as demand stays some steps ahead of supply. As the global economy rewires for the AI age while token-heavy AI applications come into their own, perhaps exponential breakthroughs justify the rise in the semi stocks.

History suggests semis boom and bust, but if you look at the token trajectory (think a potential token explosion) and the magnitude of the AI boom, maybe things really are different this time. And that warrants a different kind of thinking and perhaps shelving the history books altogether.