Markets seem to be listening to AI doubters, for now. The remarkable observation is that doubters have been loud and clear since September 2025, including AlphaValue’s modest voice. If European market observers had then a sense of unease, it must have been felt well before by US market professionals. It implies that markets are seemingly cooling down on a more than 12-month uninterrupted AI triggered speculation.
The gap between seeing correctly and acting differently is obviously down to the fast money opportunity. The partying on commissions gained on stratospheric capital raisings of all feathers found its apex with the SpaceX listing. Subsequent sycophantic research by US investment banks would make TMT bubble analysts of yore blush.
In this moment of cooling, rather obvious comments can be reiterated:
The first and most important is that debt markets are always calling the shots. We knew and wrote that Google going to the bank to match negative FCF would be a tipping point. It happened, and the initially cosy spreads on the funny SPVs aimed at pushing capex off balance sheet are fast moving into junk territories. With clarity on risk and thus less no-question-asked debt funding of an industry with no demonstrated business model, the hoovering of capital from unsuspecting retail hands will most likely be more difficult too.
The second is that Semiconductors, astounding technologies as they are, are as much in a hog cycle as they have ever been. Every half lucid investor knows that 60% quarterly price increases mean that buyers, whether rich from AI circles or poor from ordinary industries, are walking away. If AI processing demand is slowing one inch and if resulting semis demand slows down by a sliver of an inch, a crash of biblical proportions should be the natural outcome.
Of course holders of Semis stocks tend to gear themselves to the hilt as they bet the house, so that a biblical crash does not start to describe the extent of the looming correction. Say 30% if late 2024 valuations are a proper anchor.
Correction? May not be brutal
Bears have been around for a year or so and have telegraphed much of their concerns, starting with opinionated financial media. The surprise factor is thus absent as a trigger to a brutal and extensive correction. Erosion would be more like it.
Next to Semiconductors, Capital Goods deserve to return to more mundane valuations. Their spot price vs. peak of June 2026 is an 8% correction. Another 15% would be deserved. Applying hog cycle principles, the sector should not trade at 27X 2026 earnings up 19%, but rather 12x earnings up 35% or more in both 2026 and 2027, and then contract to pre AI growth dreams.
Building Materials too are Deep Cyclicals reborn under AI auspices as quasi Capital Good stories. There is nothing to justify stretching the PE from an acceptable 11-12x to a current 15-16x. As their earnings may well be driven up by a strong pricing impact that begs for a collapse in demand, we would suggest that 8x to 11X is a proper PE on peak earnings.
The same words would be applicable to the world of magnificent Miners. They used to trade at 6 to 8X. This is where they belong at peak earnings (current and driven by copper and specialist metals), not at the current 15X.