Short sellers are often treated as the vultures of the public markets, circling dying companies and waiting for the inevitable decay. But Carson Block, the founder of Muddy Waters Research, argues that we are entering an era where they function more like structural engineers identifying a fundamental flaw in a skyscraper before it collapses. The catalyst for this shift isn't just a cooling economy or a spike in interest rates. It is the widespread, often reckless adoption of generative artificial intelligence. For a short seller, AI isn't just a tool to scan spreadsheets; it is the greatest creator of corporate smoke and mirrors the financial world has ever seen.
The thesis is straightforward. Companies are currently inflating their valuations by slapping an AI label on legacy business models that the technology will actually destroy. Block suggests that the gap between corporate marketing and operational reality has never been wider. This creates a target-rich environment for those willing to bet against the hype. In the past, shorting required finding a company cooking the books or hiding a massive debt load. Today, you just have to find a CEO claiming their mid-market SaaS platform is suddenly a "machine learning powerhouse" while their actual customer base is migrating to open-source alternatives. If you found value in this article, you might want to read: this related article.
The Mirage of Productivity
The market currently rewards any mention of automation with a bump in share price. This is a mistake. Most executives view AI as a way to cut headcount and boost margins overnight. They ignore the reality that their competitors are doing the exact same thing, which leads to a race to the bottom. If every firm in an industry cuts costs by 30% using the same large language models, that savings isn't retained as profit. It is passed to the consumer through price wars. The competitive advantage vanishes.
Short sellers look for the "incumbent’s dilemma" on steroids. Think about companies that rely on high-volume, low-complexity tasks like data entry, basic legal discovery, or Tier 1 customer support. These firms are priced as if they will successfully transition into AI providers. In reality, their clients are realizing they can replace the entire firm with a custom API. The revenue doesn't just dip; it evaporates. Block’s strategy involves identifying these "zombie incumbents"—firms that are dead but haven't stopped trading yet. For another look on this event, check out the latest update from MarketWatch.
Data Moats Are Often Just Puddles
One of the biggest lies told to investors is the concept of the proprietary data moat. Companies claim their historical data makes their specific AI models untouchable. This is frequently nonsense. Much of this data is unstructured, dirty, or legally restricted. More importantly, the hardware costs required to train models on that data are astronomical.
A company might spend $500 million on compute power to build a tool that a nimble startup can replicate for a fraction of the cost six months later using more efficient architectures. The "moat" is actually a capital-intensive anchor. Investigative analysts are now digging into the depreciation schedules of these hardware investments. If a company buys thousands of GPUs but cannot turn a profit on the services they provide, they aren't a tech leader. They are a failing utility company with very expensive heaters.
The Problem of Synthetic Growth
We are seeing a rise in what can be called synthetic growth. This happens when a company uses AI to juice its metrics in ways that don't reflect actual market demand.
- Automated Lead Gen: Sales teams use bots to flood the market with "interest," creating a bloated pipeline that never converts.
- AI-Generated Content: Media and marketing firms pump out volume to maintain SEO rankings, ignoring the fact that search engines are pivoting away from rewarding low-effort bulk.
- Code Bloat: Software companies use AI to write more code faster, but the resulting technical debt makes the product unstable and impossible to maintain.
Short sellers are beginning to use their own specialized models to detect these patterns. They look for anomalies in traffic, engagement, and code commits. When they find a company whose output is increasing while its meaningful customer engagement is flatlining, they have found a short.
The Return of the Forensic Accountant
For a decade, cheap money made everyone look like a genius. High interest rates changed the math, but AI changed the storytelling. We are seeing a return to the era of forensic analysis, but the tools have evolved. Traditional short selling involved looking for "red flags" in a 10-K. Modern short selling involves verifying the "intelligence" in the AI.
If a company claims its new platform uses proprietary neural networks, a skeptical analyst will test the latency. If the response time suggests the app is just a wrapper for a basic GPT-4 call, the valuation is a fraud. There is no intellectual property. There is only a subscription to someone else’s engine. Investors are paying a premium for a middleman who is about to be disintermediated.
High Stakes and Higher Risks
Shorting in an AI-driven market is not for the faint of heart. The "AI halo effect" can keep a stock irrationally high for years. A short seller can be 100% right about a company’s eventual demise and still go broke waiting for the market to realize it. This is the danger of the "short squeeze," where retail enthusiasm or algorithmic buying drives a stock price up regardless of fundamentals.
Carson Block’s approach focuses on the "terminal value." He isn't looking for a 10% dip. He is looking for companies whose terminal value is zero. When the core product of a company becomes a commodity available for free through an open-source model, the stock price eventually follows. The trick is timing the moment the "innovation" narrative flips to a "legacy" narrative.
The Architecture of Deception
Corporate communication has become a feedback loop of buzzwords. CEOs are coached to mention "autonomous agents" and "vector databases" in every earnings call because they know it triggers algorithmic buying. This creates a massive disconnect. The people buying the stock don't understand the tech, and the people running the company don't understand the tech. They are both betting on a miracle.
The short seller’s job is to be the only person in the room who actually read the documentation. They talk to the engineers who are being told to build impossible features. They talk to the customers who are frustrated by hallucinating interfaces. They find the friction that the marketing department smoothed over with high-gloss graphics.
Identifying the First Dominoes
The first wave of AI-driven collapses will likely happen in the BPO (Business Process Outsourcing) sector. Any firm that sells "human hours" for digital tasks is in immediate danger. Why pay a firm in Manila or Bangalore to categorize invoices when an LLM can do it for $0.001? These companies have massive overhead, thousands of employees, and no way to pivot fast enough.
Next come the "SaaS Overlords." For years, companies paid $50 per user per month for software that managed basic workflows. Now, those workflows can be built in-house using AI-assisted coding tools. The "Great Unbundling" of software is starting. Every subscription fee is a target for replacement.
The End of the Generalist
The era of the generalist investment analyst is over. To short a company in 2026, you need to understand the difference between inference costs and training costs. You need to know why a specific GPU cluster is a liability rather than an asset. You need to understand that "AI" is not a singular thing, but a broad category of technologies, most of which are rapidly becoming commodities.
The most successful short sellers are hiring data scientists not to build trading bots, but to stress-test the claims of the companies they are betting against. They are running their own simulations to see if a company’s "breakthrough" is actually just a standard feature of the next software update from a big tech provider.
A Brutal Reckoning
Markets eventually revert to the mean. The excitement surrounding AI is real, but the financial structures built on top of that excitement are often flimsy. We are currently in the "over-estimation" phase of the technology cycle. Investors assume every company will find a way to monetize AI effectively. History tells us that only a handful will, while the rest will spend their remaining cash chasing a ghost.
Short sellers like Block aren't betting against progress. They are betting against incompetence and the human tendency to believe in magic during times of rapid change. The "new dawn" for short sellers isn't about a market crash. It is about a permanent shift in how we value a business. In a world where machines can generate content, code, and strategy, the only thing that has value is what a machine cannot do. Everything else is a short.
Look at the companies currently trading at 50 times earnings while their core product is being integrated into the operating system of the devices their customers use. That isn't growth. That is a liquidation sale disguised as a tech revolution.
