In the wake of Samsung’s earnings report and, cutting to the chase, a selloff in the locally-listed shares despite a 19-fold increase in operating profit, it’s worth delineating the “problem.”

Because when you grow the bottom line 1,800% and the stock falls, there’s a problem somewhere, even if you can chalk up the knee-jerk reaction to unrealistic expectations and retail investor leverage.

This’ll seem repetitive, so I won’t belabor the point beyond what’s necessary to (re)communicate it.

Investors are worried that the hyper-scalers, having seen enough in the way of underperformance from their own stocks, may soon tip an inclination to scale back capex. Reports that Meta’s considering a business to sell excess compute were read by some as a harbinger in that regard.

Even if that’s not the case — i.e., even if outright capex cuts / spending guidedowns aren’t on the cards — the growth rate‘s going to slow. That much is a foregone conclusion.

The figure above, from SocGen’s Manish Kabra, gives you a sense of things.

The concern for semi stocks is this: When hyper-scaler capex growth starts to roll over, so too will forward profits and/or revisions breadth for chips. At that point, the parabolic rally in the stocks will lose its fundamental underpinning.

An equal-weighted SocGen basket of 46 so-called “AI beneficiaries” is up 70% this year. The hyper-scalers (i.e., an equal-weighted index of Alphabet, Amazon, Meta, Microsoft and Oracle) are down 7%.

That’s a 75ppt performance disparity, and it’s unlikely to prove sustainable. It doesn’t have to “resolve” with a semi crash, but it sure as hell isn’t going to square up via a 90-degree-angle rally in five stocks with market caps the size of the hyper-scalers’. That’s not mathematically realistic.

Given that, some (many) believe chips are due for a correction. And it’s becoming more difficult to argue otherwise in the face of myriad distortions.

The figure on the left, below, from the same SocGen slide deck, shows you semis are expected to account for more than two-thirds of S&P EPS growth this year and well more than half in 2027.

The figure on the right shows you that as a share of overall market cap, semis are on the verge of eclipsing the hyper-scalers. That, to me, is a stretch. The hyper-scalers have diversified business models, notwithstanding the “eggs in one basket” risk they’re taking with massive AI outlays.

There’ll come a point, probably sooner rather than later, when management says, “We need to take a breather and let this play out for a few quarters, because investors are plainly anxious and we just don’t know the ROI on this spending yet.” Maybe they won’t say that aloud, but it’ll be there in the subtext.

When that moment comes, semis could look a little like Wile E. Coyote. And the whole “this time’s different” narrative re: the cyclicality of, for example, the memory business, will be judged the same way (almost) all “this time’s different” narratives are eventually judged.