In short: Cautious on the equities, not the earnings: "I tend to be more cautious about the equities there." AI is "the sixth big capex boom in US history," the biggest as a share of GDP; history says it can run, but two to three years in "it paid to take some off the table and just put it in gold because gold actually outperformed the capex boom sector over the full course of the rest of the cycle." Concedes Dale is "exactly right" that earnings keep improving.
This is Gromen's caution on the AI stocks, not a bet against them. He agrees with Dale that AI earnings will probably keep improving, and he likes the physical build-out. His hesitation is historical: AI is the sixth great capital-spending boom in US history, after canals, railroads, electrification, highways and telecom, and the largest as a share of the economy.
His firm's study of those booms found they can run for a while, but two to three years in — with stock valuations as stretched as they are today — long-term investors did better trimming and moving the proceeds into gold, which beat the boom sector over the rest of every cycle. He also flags a less obvious risk: AI is being built to replace workers, and about half of federal tax revenue comes from employment, so AI's success can undermine the government's finances.
1:25:14And that's where I get it starts to sort of hit this singularity or this asymptotic. I don't know how quite to think of it and how those factors will interplay. I tend to be more cautious about the equities there and in part that is it's we did some work a couple weeks ago that highlighted if you AI is the I think the sixth big capex boom in US history.
In short: Take profits, don't short. "The bubble's not over yet" — the SEC loosening securitization rules for AI paper and Jensen Huang's compute-derivative talk are "making it easier to get more credit to them," and new credit is exactly what keeps a mania running. But: "once you were two to three years into any of those other bubbles… you did better by selling most of the bubble and buying gold. Gold outperformed over the full cycle." Explicitly not a short: "I wouldn't short them here."
His position on AI chips is deliberately two-sided: take profits, but don't short them. "I wouldn't short them here."
The reason the bubble isn't over is credit. Reading about the South Sea Bubble and John Law's Mississippi scheme, his takeaway was that these episodes only end when the new credit stops arriving — and right now it is arriving in volume: the SEC is loosening the rules on packaging AI-related loans into securities, private equity firms are being convened to do exactly that, and NVIDIA's CEO is publicly floating a compute-based derivative. "No judgment. I know what that is. That's making it easier to get more credit to them."
The other side is the timing rule from his own research. Across every previous capital-spending boom going back to the 1840s canals, an investor who was two or three years in did better by selling most of the position and buying gold — gold won over the full cycle every single time, even when the boom itself ran on for a while. So the stance is: stay in for now, trim as it runs, and know what you're rotating into.
1:11:35And I think this is a part that people aren't paying enough attention to. And I'm not saying this is — again I agree with you, I wouldn't short them here. When I say take profits, the analysis in that report that you have shows that once you were two to three years into any of those other bubbles, you did better if you were a long-term investor, you did better by selling most of the bubble and buying gold.
In short: Nuanced: "gun to my head" the rally has been too much too fast; if US AI has a problem, semis get dragged down with it — "probably an opportunity to add." But China's "cheaper and better" AI moment (Z.AI's NVIDIA-free 1GW data center) means within ~5 years the multiple you'll pay for semis has to fall. Near-term down, "ultimately probably higher."
Here Gromen is deliberately two-handed. Near term, he thinks the semiconductor rally has run "too much too fast," and if the US AI story stumbles, chips get sold off with it — which he'd treat as a chance to add, not a reason to flee. Longer term he's more cautious: he's watched China go from "cheaper but worse" to "cheaper and better" in industry after industry, and it's now happening in AI (a Chinese firm just ran a huge data center on Chinese-only chips). If China does to semiconductors what it did to solar and EVs, investors will eventually pay a lower multiple for them. So: probably higher in price ultimately, but on a shrinking valuation — a genuine hold, not a clean buy or sell.
40:42But so for me, I look and go near term, I think AI chips get dragged down with AI, on that risk, but ultimately I think they probably end up higher ultimately. Seems to me that long-term China has the energy advantage. They have the strategic commitment to AI and they're not backing down.
Nothing matches this filter.
Verbatim excerpts from the public transcripts (auto-pulled at each mention's timestamp, lightly cleaned). Timestamps link into the video; "source page" opens that commentator's full analysis at this row. Click a mention's header line to expand it (one open at a time). For personal study — not investment advice.