Every so often it’s worth asking: what’s overrated right now, what’s underrated, what’s been forgotten. But don’t judge by price — price is only the outcome. The measuring stick is narrative density: how far the story an asset tells has drifted from the part of it that actually works.
Overrated
Four things, sharing one trait: far more story than substance.
Pure-concept AI names. All story, no execution — no different from .com companies. Twenty-five years ago they died from having no cash flow, not from having no imagination.
Moatless AI upstream segments. Optical modules, data centers, commodity compute servers — capacity is being added far beyond long-term demand, and in two or three years the glut is inevitable. The moatless segment is the first thing competition eats.
AI commercialization expectations. What the market overrates is the speed of adoption and the efficiency of monetization. The story is about the endgame; the P&L is about next year.
Valuations of the top AI chip giants. Pricing in ten straight years of high growth from now, when the history of semiconductors is precisely a history of cycles. There has never been a “forever.”
Underrated
Four things, sharing one trait: the fundamentals are fine, nobody is just looking.
Old growth sectors drained by AI. Biotech, renewables, advanced manufacturing: fundamentals intact but capital keeps flowing out, valuations pressed to historic lows. Real people are still doing real work in these old sectors; they didn’t do anything wrong — the hot money just changed direction.
Traditional resources and cyclical commodities. Energy, mining, agriculture. Narrative density near zero, demand nowhere near zero.
Stable-cash-flow value assets. Utilities, consumer staples, offline services — pushed out of view by the “tech narrative above all” reflex. Nobody talks about liquor stocks anymore; the market’s hottest story is now tech. Every time attention moves house, valuations get reshuffled.
Crypto and Web3. Retail is gone; the bottom is where nobody looks. Bitcoin itself didn’t disappear — its narrative priority dropped; the rotation itself was dissected in “Silence Is What a Bottom Sounds Like”, so only the conclusion is borrowed here.
The Most Overlooked Item
Not anything on either list, but the lag in overcapacity. The frantically expanding AI infrastructure comes online one to two years from now, all at once; by the time the glut is visible, the price has already finished its move. The lag is exactly why nobody sees it coming — and that is what makes it dangerous — the lag mechanism itself was laid open in “The Bubble Died, the Rails Remained”; this piece cites it as a known quantity.
The other side deserves its hearing: when do the underrated come back? The honest answer is that there is no date. A bottom is a phase, not a day on the calendar; underrated can get more underrated, overrated more overrated. And the judgment “AI adoption is too slow” may itself be wrong — adoption may genuinely be fast, monetization may genuinely work. That is not a flaw in this piece; that is its falsifiable edge. Look back in two years, and the numbers will say who was right.
So the lists answer position, not action. Overrated and underrated is an attention problem, not a value problem — value didn’t evaporate overnight; attention moved. Where attention left is not automatically a trap, and where attention piles up is not automatically gold. A position judgment is not a buy or sell signal, and neither list is a recommendation of anything. The lists answer one question only: right now, where is everyone looking?
