The "smart money" is currently obsessed with hunting for the next Enron in the silicon valleys of the world. You’ve seen the headlines. Short sellers are salivating over companies that slapped an AI label on a basic spreadsheet and watched their stock price double. They call these "fake AI stocks." They think they’ve found an easy arbitrage opportunity by betting against companies with high price-to-earnings ratios and low actual compute power.
They are wrong. They are fundamentally misunderstanding how technical debt and market psychology operate in a speculative bubble.
If you are looking for "fake" companies to short, you are playing a game that ended in 2001. In the current market, being "fake" doesn't mean a company will fail. It often means they have lower overhead and higher agility than the giants building the actual infrastructure. Shorting a bubble isn't about being right; it’s about timing. And if you’re timing based on technical purity, you’re going to get liquidated.
The Myth of the Technical Moat
The common argument suggests that if a company doesn't own its own Large Language Models (LLMs) or doesn't have a proprietary chip architecture, it’s a fraud. This is the first "lazy consensus" that needs to die.
Ownership of the stack is actually a liability right now. The rate of depreciation on AI hardware is staggering. If you bought a massive cluster of H100s six months ago, you are already sitting on aging assets. Meanwhile, the "fake" AI company—the one that just builds a sleek interface on top of an API—has zero capital expenditure on hardware. They are the ultimate scavengers.
I have seen companies burn $50 million trying to train a "sovereign" model only to have a three-person startup using a generic open-source model outperform them in three weeks. The startup is the one the short sellers are targeting because it lacks "depth." In reality, that lack of depth is why they can pivot while the giant is still trying to figure out why their weights are drifting.
The market doesn't pay for the plumbing. It pays for the faucet.
Why "Vaporware" Outperforms "Hardware"
Short sellers love to point at companies that have no revenue but a massive valuation. They call it a bubble. It is a bubble. But bubbles last decades, and they don't pop because a company is "fake." They pop when the liquidity dries up.
Current short-seller logic:
- Company X says they do AI.
- Company X actually just uses a wrapper for a GPT model.
- Therefore, Company X is worth zero.
This ignores the history of every major technological shift. In the 90s, the "real" companies were the ones laying the fiber optic cable. The "fake" companies were the ones building websites that sold pet food or books. The fiber companies went bankrupt. The "fake" website companies became the rulers of the modern world.
Right now, the "real" AI companies are the ones spending billions on electricity and cooling. They are the infrastructure. The "fake" companies are the ones figuring out how to actually make a human being pay $20 a month for a tool.
If you short the "wrapper" companies, you are betting against the only part of the ecosystem that actually touches the end consumer. That is a losing bet.
The Valuation Trap
The skeptics point to $P/E$ ratios. They see a company trading at $100\times$ earnings and think they’ve found a gift.
$$P/E = \frac{\text{Price per Share}}{\text{Earnings per Share}}$$
When earnings are near zero, the ratio becomes a useless, infinite number. Shorting based on $P/E$ in a growth sector is like trying to measure the speed of a jet with a ruler. It’s the wrong tool for the job.
The real metric you should be looking at is the cost of customer acquisition (CAC) relative to the lifetime value (LTV). Most of these "fake" AI companies have a CAC that is effectively zero because of the hype. They are getting free marketing from the very short sellers who are complaining about them. Every time a skeptic writes an article about how "overvalued" a niche AI video editor is, they are providing that company with millions of dollars in free brand awareness.
The "Intellectual Property" Delusion
"But they don't own any IP!" the short seller screams into the void.
In the world of software, IP is mostly a myth used to soothe VCs. The "moat" isn't the code. The moat is the workflow integration. Once a company embeds itself into the daily habit of a worker, it doesn't matter if the underlying engine is "fake" or borrowed.
Imagine a scenario where a legal firm adopts an AI tool to summarize depositions. The tool is just a wrapper on an LLM. But the tool integrates with their billing software, their file storage, and their email. If a "real" AI company comes along with a slightly better model but no integrations, the law firm will never switch. The friction of moving is higher than the value of the "better" technology.
Short sellers are looking at the engine; they should be looking at the dashboard.
The Danger of Being Right Too Early
The most dangerous thing in finance is being right at the wrong time. Most of the companies being called "AI frauds" today will indeed be gone in ten years. But between now and their demise, they could easily see their stock prices rise by another $400%$.
Shorting is a game of infinite risk. When you buy a stock, your downside is $100%$. When you short a stock, your downside is theoretically infinite. Betting against a cult-like tech trend because you think you’re smarter than the "retail plebs" is the ultimate ego trap.
I’ve watched funds get incinerated because they tried to short Tesla or Netflix based on "fundamentals." The fundamental they missed was that the market can remain irrational longer than you can remain solvent. In AI, the "irrationality" is actually a rational bet on the total restructuring of human labor. Even if $90%$ of the players are faking it, the $10%$ that aren't will carry the entire sector's valuation to the moon.
How to Actually Spot a Loser
If you insist on betting against the AI trend, stop looking at the technology. Stop looking at the "fakeness." Start looking at the burn rate versus the "hype cycle" duration.
The losers aren't the companies with "fake" tech. The losers are the companies that are trying to build "real" tech without the capital to sustain it. The companies trying to compete with Google and Microsoft on a base-model level are the ones headed for a cliff.
Look for the companies that are hiring thousands of engineers to build something that will be a commodity feature in Windows 12 next year. That is where the real fraud lives. It’s not the startup with a clever wrapper; it’s the mid-cap legacy player trying to reinvent the wheel using $1990\text{s}$ management structures.
The Truth About "Fake" AI
The term "AI" itself is a marketing term, not a technical one. We are dealing with advanced statistical modeling and probabilistic inference. There is no "real" AI in the sense of a conscious machine. Therefore, everyone is "faking it" to some degree.
The short sellers who think they can distinguish between a "pure" AI company and a "fake" one are usually just showing their own technical illiteracy. They are looking for a binary distinction in a spectrum of implementation.
If a company uses AI to automate a process that was previously manual, and that automation creates value, it doesn't matter if they used a "real" proprietary neural network or a series of complex "if-then" statements. The market pays for the result, not the method.
Stop Looking for Frauds and Start Looking for Friction
The real opportunity isn't in shorting the high-flyers. It’s in identifying the "boring" companies that are going to be crushed because they refuse to use the "fake" tools.
The danger isn't the bubble. The danger is the transition. Companies that are "pure" and "authentic" and "human-driven" are the ones that will be liquidated. They are the ones with the high costs and the slow turnaround times.
The "fake" AI stock—the one with the high valuation, the thin tech, and the aggressive marketing—is likely to survive because they have the one thing that matters in a gold rush: a shovel that people actually want to use, even if it’s made of plastic.
If you’re betting against the plastic shovels because you’re waiting for someone to invent a teleportation device, you’re going to die in the desert.
Stop trying to be the smartest person in the room. The smartest people in the room are the ones currently making a fortune selling "fake" AI to people who don't know the difference, and the second smartest are the ones buying the stock because they know the hype train has no brakes.
Short the laggards. Long the "frauds." Stay solvent.
