A corollary of the wild surge in semiconductor stocks is an equally enthusiastic bull market in statistics and charts documenting various manifestations of semi dominance.
If the “mania,” as JonesTrading’s Mike O’Rourke described the fitful exuberance engulfing chip stocks in April, feels like a recent phenomenon to you — an “all of a sudden” sort of thing — it’s not your imagination.
The overarching bull case isn’t new, of course. Neither is the notion that in a geopolitical environment defined increasingly by competition instead of cooperation, chips are akin to scarce resources. And we’ve certainly seen parabolic semi rallies before.
What’s new is the near-vertical infection across the entire space.
When considered atop the triple-digit gain from the “Liberation Day” lows last year, the one-month SOX rally from the Iran war dip is steep enough to render all but the shortest lookback linear charts useless. You really need a log scale now.
“In many respects, the narrative for global equity markets this year has been disarmingly simple,” SocGen’s Andrew Lapthorne said Monday.
That narrative: The largest, most dominant companies in the history of capitalism have committed to plow what, ultimately, will be trillions of dollars into AI compute, an unprecedented capex spree that’s now manifesting as demand.
“Semiconductors sit at the very center of this theme, with tech hardware close behind,” Lapthorne said, noting that semis now account for around 15% of the MSCI All-Country World Index. For context, that share was — get this — just 3% at the beginning of 2019.
This isn’t entirely (or even mostly) without justification. As Lapthorne put it, it’s “hard to dismiss the fundamental momentum underpinning the move.”
The figure on the left, above, suggests we’re currently witnessing the largest rolling four-month upswing in global EPS expectations ever, and the biggest percentage increase on record over a four-month window outside of rebounds from recessions.
“12-month forward profits are now around $600 billion higher than they were just four months ago,” Lapthorne went on, in the same note.
Between them, semis, IT hardware and energy account for 98% of the increase in 2026 consensus profit forecasts, as illustrated by the chart on the right above. (You can read more about that here.)
This raises all-too-familiar concerns. With semis and tech hardware responsible for nearly half of global equities’ gains in 2026, “concentration risk” is front and center “once again,” Lapthorne said.
You referred to the Kospi earlier today and specifically pointed to SKHynix. That firm dominates the memory chip market along with Micron Technology.
Historically memory has been a notoriously cyclical market which usually punishes producers who expand at or close to the top of the demand cycle.
Well, you’re assuming we’re “at or close to the top of the demand cycle.” I think a lot of people — not me, necessarily, but a lot of people — would argue we’re nowhere near that. As noted here over the weekend, the historical relationship between semis, broadly, and ISM manufacturing, for example, suggests the latter’s on its way to >60. And fwd EPS estimates for the S&P, which are of course heavily skewed by semis, tech and, now, energy, suggests NFP payrolls are about to spike sharply (maybe not next week, but in the months ahead). In other words, you’re assuming an end-of-cycle environment (or seem to be) and a lot of other people are thinking the exact opposite.
Fortunately for humanity, a lot of people are smarter than me. I thank them for running up the value of some of my all too meagre chip holdings which I have slowly been trimming over the last couple of weeks.
If autonomous driving and robotics materialize in the near future, then their need for compute will be enormous. The voracious demand for DRAM, NAND and inference driven low-latency SRAM, will dwarf the notoriously vertiginous cyclical history of memory.
That is what Elon Musk (who is smarter than me) is saying. At the moment, self-driving cars must be tethered to a cloud. Witness the traffic jams they caused in San Francisco when there was a temporary power blackout. No doubt some state actors noticed that as well.
But Elon’s believers mostly seem to be unaware that robots are NOT some new game changer. I’ve been investing in robotics and automation for over 15 years so I’ve followed them pretty closely. They are used all over the globe, including cobots which are already widely used on factory floors, working side-by-side with humans without the need for AI. Musk is betting the ranch on designing and producing cobots in a human form, adding hands with improving dexterity. But his forecast numbers are pretty damn fanciful. There is little incentive for the adoption of humanoid robots on factory floors. They may prove useful in some warehouse and distribution roles which might benefit from their promised dexterity. But I’d argue that the TAM is not as large as Musk thinks. The killer use would be in eldercare where it’s unclear how many will be sold with a $700,000 price tag.
Pass the bong, turn up the music and order more pizza.
Maybe this is already obvious to everyone else, but isn’t the recent face ripping run-up in semis a direct result of the closing/blockade of the Strait of Hormuz? Helium is one of the main exports that comes out of the Strait, and it is essential to the production of semiconductor chips. If you are going to build hundreds of billions of dollars worth of data centers across the U.S., and future chip production may be slowed dramatically, you had better get your company’s hands on as many chips as you possibly can (by front-ordering) ASAP. It is war speculation — and potentially another card Iran now holds — and it could all reverse very quickly once the Strait is re-opened, no?
Lesson learned from Covid supply chain disruption.
I should also add that a supply of new chips is essential to the hardware companies that will be making the machinery other industries will need to upgrade their systems in order to maximize their investments in AI.