The tell in an AI bubble is not the valuation of the leaders. It is the quality of the companies arriving late to the party. When a business whose only ...
The tell in an AI bubble is not the valuation of the leaders. It is the quality of the companies arriving late to the party. When a business whose only route to public markets runs through a reverse merger into a dead operating shell suddenly appears on the micro-cap conference circuit wearing a fresh AI narrative, the market is broadcasting its own position on the quality curve. That signal is more reliable than any multiple on the megacaps, because it measures something the leaders cannot: how far down the promoter ladder capital has been forced to reach to find a new story to sell.
Why the Late Arrivals Are the Barometer
Quality of companies coming public tends to move inversely to how stretched a bull market is. Early in a cycle, the businesses that access public capital are the ones that can. They have real revenue, a defensible product, an underwriter willing to attach its name. As prices climb and the obvious winners get expensive, the marginal issuer degrades. The story stays the same; the substance behind it thins.
Ted O'Glove, who wrote Quality of Earnings, spent a career teaching investors to separate reported profitability from the money a company actually makes. The framework he built was defensive by design: it assumed that the further you got from cash, the more room there was for a narrative to substitute for a business. The reverse-merger shell is that principle at its logical extreme. There is no operating business to analyze. There is a shell, a ticker, a deck, and a theme borrowed from whatever is working that quarter. In 2014 the theme was often China. Today it is AI.
The mechanism is not that these companies cause the top. They do not move enough capital to matter. The mechanism is selection. A promoter with a marginal asset and a choice of narratives picks the one with the most credulous buyers. When AI is the chosen wrapper for the worst available deals, it tells you the credulity is concentrated there. The barometer reads the promoter's revealed preference, and the promoter is a professional at reading the crowd.
The Reinvention Pattern
The item that prompted this is not the shell itself but the histories behind several of the larger AI winners: companies that reinvented themselves once, twice, sometimes more, each time latching onto the money grab of the moment before rebranding into the next. That pattern is the same signal viewed from the other end of the quality scale. A company that was a crypto play, then a metaverse play, then an AI play has told you it is theme-driven rather than product-driven. The through-line is not a technology. It is an instinct for where retail attention is pooling.
This is classically what bubbles produce. Not fraud in every case, though some of it is fraud, but a steady migration of the marginal operator toward the hottest label. The reverse merger is the vehicle because it is the fastest, cheapest, least-scrutinized route to a public quote. No roadshow gate, no underwriter reputation on the line, no lockup discipline. A shell that never amounted to more than a shell absorbs a story and starts appearing on the small-cap circuit with a deck engineered for the moment.
The defunct Chinese reverse mergers of the last cycle are the reference class. They arrived the same way, with the same slickness, into the same appetite. Most of them are gone. That history does not prove every AI-labeled shell fails, but it establishes the base rate for the vehicle, and the base rate is unkind.
What Would Break This Read
The honest counterargument is that the signal has almost no predictive precision on timing. Junky reverse mergers have appeared in AI-labeled form for a while now, and the leaders have kept climbing. A barometer that tells you the air is thin does not tell you when the descent begins, and markets can stay in the "stupid is as stupid does" phase far longer than the skeptic's patience. Reading the shells as a top-caller is a category error. They mark the phase, not the date.
There is a second, harder objection. Some genuine technology waves also draw opportunists early, and the presence of junk does not falsify the underlying platform. The internet bubble minted both Pets.com and Amazon. AI can be a real productivity shift and simultaneously the wrapper for the year's worst deals. The shells being garbage says nothing definitive about whether the leaders are overvalued. That link is the one to hedge: the barometer measures crowd credulity and promoter behavior, not the fundamental value of the megacaps. It is evidence about the phase, not a valuation verdict on any specific name.
The Berkshire thread that opens the original item carries the same discipline in reverse. The argument that a conglomerate has grown too complex to compound at its old rate is a quality-of-earnings argument, not a fraud accusation. Nobody is too smart to fail, and being too smart by half is its own hazard, as Long-Term Capital Management and Enron each demonstrated in their own register. But complexity dragging on returns and a shell borrowing a narrative are different failures. One is the erosion of a genuine machine; the other is the absence of a machine behind a story. Conflating them weakens both arguments.
The Condition to Watch
The read confirms itself if the flow accelerates: more reverse mergers into empty shells, more serial-reinventors adopting the AI label, more slick decks on the micro-cap circuit built around businesses with no cash generation to speak of. Quality of the issuance stream is the observable. When the marginal AI issuer gets worse and the story gets louder, the phase is deepening, whatever the leaders are doing.
The read weakens if the opposite shows up: if new AI issuers increasingly carry real revenue, real customers, real cash, and if the serial-reinventors stop finding buyers for the next rebrand. That would signal the credulity is draining rather than concentrating, and the barometer would be pointing the other way.
None of this is a trade. It is a way of reading where you are. The leaders' multiples tell you what consensus believes about the future. The garbage arriving at the back door tells you how much consensus is willing to believe, and from whom. In a bubble, the second number is the one that turns first.
