Any market skeptic worth his or her salt will tell you this time’s never different. Every boom and bust cycle’s the same from a psychological perspective, they’ll say.

Excitable investors caught up in the latest frenzy spin yarns that purport to explain why the old “rules” don’t apply, and although some such narratives will sound plausible enough in the moment, hindsight will show them to be every bit as injudicious as the fanciful stories conjured to excuse past bubbles which eventually burst.

Currently, some argue that the structural shortages accompanying the AI buildout mean this time’s different for memory stocks. I discussed this at some length last week while documenting the absurdist rally in shares of SK Hynix which, despite having surged more than 1,000% over the past 12 months, still trade at a bargain-basement multiple.

“It’s easy enough to lampoon the stock’s vertical ascent as absurd,” I wrote, on the way to sketching the broad contours of the argument:

The bull case says this cycle’s different for the memory chip industry. That AI’s changed the game. That, at the very least, the boom phase can last much longer this time. Maybe even “forever,” given the severity of the shortage and the size of the implicit backlog.

I don’t necessarily believe that, but I don’t necessarily disbelieve believe it either.

The is a pretty vexing dilemma for market participants in the current environment. The big three memory stocks (Micron, Samsung and SK Hynix) have experienced parabolic rallies and considering nothing’s more “fundamental” than supply and demand, it’s difficult to make the case the gains are purely speculative. Demand’s insatiable, and supply’s severely short.

Then again, memory’s absolutely notorious for boom-bust cycles. Is this time really different? Only if AI is in fact a world-changing innovation on par with history’s greatest technological epochs, and even if it is, the market for memory will balance eventually. The question is about the time line on “eventually.”

“Compositionally, the Memory stocks driving a large share of S&P 500 earnings growth this year typically trade at low valuation multiples due to the cyclicality of their earnings,” Goldman’s Ben Snider wrote, remarking on the quandary.

The figure above gives you some context for the discussion. Plainly, investors won’t pay as much for what they suspect is but a fleeting earnings jolt as they will for developments seen as long-term game-changers.

“Some investors,” Snider went on, “believe the AI investment boom is likely to last long enough that the associated earnings should be treated as secular, rather than cyclical, profit streams.”

That’s the crux of the issue. And, as Snider was keen to emphasize, “enough investors remain skeptical” of that thesis to make “a major increase in valuation multiples unlikely” in the near-term.

Now consider the figure below, which shows you the estimated share of US equity value attributable to long-horizon prospects.

That’s basic B-school stuff, but it’s a helpful reminder: Generally speaking, anywhere between two-thirds and three-quarters of equity valuation can be explained by the estimated worth of the company as a long-term, going concern.

What counts, as Snider put it, is “earnings far in the future.” Right now, it’s quite difficult, which is to say impossible, to game that out in the context of AI’s impact on corporate balance sheets.

“The biggest question for the valuations of most stocks [currently] is the impact of AI on long-term earnings growth,” Snider said, summing it all up. “But that question is unlikely to be resolved any time soon.”