Some short sellers are seeing opportunity in this tech mania. How they're spotting fake AI stocks
By Maksym Misichenko · CNBC ·
By Maksym Misichenko · CNBC ·
What AI agents think about this news
The panel generally agrees that the AI sector faces significant risks, with the most pressing being the potential for a 'grey market' shadow supply chain to erode Nvidia's pricing power and the long-term threat of power shortages for AI data centers. However, there's no consensus on the timeline or severity of these risks.
Risk: The emergence of a grey market shadow supply chain that could compress Nvidia's ASPs within the next 12 months.
Opportunity: The potential for true platform winners like Nvidia and AMD to power a multi-year rally despite some darlings crashing.
This analysis is generated by the StockScreener pipeline — four leading LLMs (Claude, GPT, Gemini, Grok) receive identical prompts with built-in anti-hallucination guards. Read methodology →
Short sellers are increasingly hunting for cracks beneath the stock market's artificial-intelligence frenzy, betting that some of the speculative excesses, copycat "AI" branding and vulnerable legacy business models could eventually unravel.
As billions of dollars flood into data centers, semiconductors and AI software, some short sellers argue the rally is beginning to resemble previous speculative manias, where weaker companies rushed to attach themselves to the hottest market theme in hopes of attracting capital and retail traders.
"A rising tide lifts all boats, and a twisting tide takes down a lot of names in the same neighborhood," Joyce Meng, founder of Fact Capital, said during a panel discussion at Sohn Investment Conference this week in New York. "Especially in the market where you have an AI frenzy, everyone trying to go jump into that, one of our favorite themes is fake AI."
Meng said she likes to run screens to identify companies that abruptly rebranded themselves to capitalize on the boom, including firms that suddenly changed their names to include the word "AI."
One target that Meng identified using the "AI name change" screen is Rezolve AI, which changed its name from Rezolve Group Limited in 2023. After digging deeper into the company, Meng said she saw multiple red flags around the business and predicted the stock to fall 60%.
Meng also pointed to a Chinese landscaping company that later reinvented itself as an AI server business. During her firm's research, she said the company appeared to have photoshopped products into marketing materials on its website and claimed to have hired employees listed on LinkedIn that turned out, according to Fact Capital's checks, to not actually work there.
The examples echo some of the increasingly surreal corporate pivots emerging during the AI boom. Allbirds, the struggling shoemaker, said last month it would rebrand itself as "NewBird AI" and shift toward compute infrastructure. The stock initially surged 582% following the announcement powered by massive retail flows before giving back most of those gains within weeks.
The Allbirds initial surge and the overall jump in stocks shows what these short sellers are up against and why their numbers have dwindled as this bull market marches on. They get their name because they borrow stock and then sell those shares, in the hopes of buying back at lower prices and returning them, capturing the difference. If a name moves higher, it can force them to buy back the stocks in order to avoid big losses.
"Trying to find more excess, where people are claiming they have it but they actually don't — for us, that's a really rich ideation opportunity," Meng said.
Fact Capital has generated positive returns from short positions since launching in 2019. Meng said she likes pairing speculative "fake AI" shorts with secular decliners across the technology industry that tend to be less volatile. She also highlighted business-process outsourcing firms and contact-center operators, particularly in India, as areas potentially vulnerable to AI disruption.
Rezolve AI declined to comment. The company reported $60 million in first-quarter revenue, surpassing its total revenue for all of 2025.
Some bearish investors are beginning to directly challenge the market's biggest winners. Culper Research disclosed a short position Wednesday in Nvidia, arguing the chipmaker faces underappreciated risks tied to China exposure.
"We recognize the stakes. Nvidia holds the single largest market capitalization on the planet, while CEO Jensen Huang has been celebrated as a generationally talented operator," Culper wrote in its report. "We are short Nvidia for one reason: the company has a significant China problem."
The short seller alleged that despite U.S. export restrictions imposed in April 2025, more than 20% of Nvidia's fiscal 2026 compute revenue remained tied to China through illegal GPU diversion and intermediaries in Southeast Asia. Nvidia has publicly said its China business effectively dropped to zero following the restrictions.
Nvidia didn't immediately respond to CNBC's request for comment.
Still, short selling in a bull market is no easy task. Major U.S. stock indexes have repeatedly climbed to record highs despite the ongoing war in the Middle East and broader macroeconomic uncertainty, as investors continue pouring money into semi makers and megacap companies tied to the AI boom.
These short sellers joined Michael Burry, who has emerged as one of Wall Street's most vocal AI skeptics. The famed investor recently warned that investors should "reject greed" and for any stocks going parabolic "reduce positions almost entirely."
Many are drawing parallels between today's AI-driven rally and the speculative excesses that preceded the collapse of many internet stocks during the dotcom era. Blue Orca Capital CIO Soren Aandahl said investors often confuse transformative technologies with guaranteed investment success.
"Railroads changed the world. The internet changed the world," Aandahl said at the panel moderated by Jim Chanos. "But many of the early purveyors of these technologies went completely bust."
Chanos, one of Wall Street's best-known short sellers, pointed to the dot-com era as a cautionary example. Chanos said U.S. economic growth and corporate profit growth in the decade following Netscape's 1995 debut were little changed from the prior decade despite the internet's transformative impact.
"There's no doubt the internet changed many, many things," Chanos said. "It didn't have a super huge impact" on aggregate economic growth.
Netscape, a pioneering web browser, was one of the defining symbols of the dot-com bubble before being acquired by AOL in 1999.
Four leading AI models discuss this article
"The market is currently pricing in perfect execution for AI infrastructure, leaving zero margin for error if enterprise adoption or hyperscaler ROI disappoints."
The article highlights a classic late-cycle phenomenon: 'thematic pivot' desperation. While shorting 'fake AI' companies like the mentioned landscaping firm is a high-alpha strategy, the focus on Nvidia’s China exposure via Culper Research is a more dangerous game. Nvidia’s valuation is predicated on hyperscaler capex, not just China-linked revenue. If AI ROI (Return on Investment) fails to materialize for Microsoft or Google, the entire sector re-rates, regardless of individual 'fake' company fraud. The real risk isn't just the 'fake' companies; it's the systemic over-leverage in the semiconductor supply chain. Shorting the 'fake AI' names is tactical; shorting the infrastructure leaders is a macro bet on a broader AI bubble burst.
The 'fake AI' pivot is often a sign of a market top, but historically, the true winners (like Amazon in 1999) survive the carnage and grow into their valuations, making shorting the broader sector a recipe for bankruptcy.
"Shorting blatant rebrands like Rezolve AI offers high-conviction downside (60% potential per Fact Capital) as AI hype discriminates winners from pretenders."
Short sellers targeting 'fake AI' like Rezolve AI (RZLV)—which rebranded in 2023 and reported $60M Q1 revenue exceeding its full prior year—spot real red flags: abrupt name changes, photoshopped marketing, fake LinkedIn hires. Allbirds (BIRD) 'NewBird AI' pivot surged 582% on retail hype before collapsing, classic momentum trap. Pairing with AI-threatened Indian BPOs is clever for lower volatility. Nvidia (NVDA) China short from Culper ignores NVDA's Blackwell ramp and ex-China dominance (China was ~20% pre-restrictions, now mitigated). Dot-com echoes overdone—AI capex is $1T+ committed, not vaporware.
Even flimsy AI rebrands could extend the rally far longer amid FOMO and retail inflows, as Allbirds' spike showed, crushing shorts before fundamentals bite.
"The real short thesis isn't whether AI is transformative—it is—but whether current valuations leave room for disappointment, and the article provides no valuation anchor to test that claim."
The article conflates two distinct problems: (1) obvious fraud/rebranding gimmicks like Allbirds' 582% pump-and-dump, which are real but already priced into penny stocks, and (2) legitimate AI infrastructure plays like Nvidia, where the short case rests on unverified claims about China diversion. The Culper Research allegation that 20%+ of Nvidia's compute revenue escapes U.S. sanctions 'through illegal GPU diversion' is presented as fact but requires extraordinary evidence—Nvidia's audit trail, customs data, and SEC filings would flag this. The dotcom parallel is seductive but incomplete: railroads and the internet eventually created massive shareholder wealth despite early busts. The real risk isn't AI transformation; it's valuation. At current multiples, Nvidia (NVDA) prices in near-perfect execution. But the article doesn't quantify what goes wrong—just that it *could*.
Short sellers have been wrong on mega-cap AI plays for 18+ months while NVDA compounded 200%+; survivorship bias means we hear from winners like Burry but not the hundreds who shorted too early and capitulated. The 'fake AI' screen catches obvious frauds, but those aren't systemic risk—they're volatility for retail traders.
"The AI rally will be sustained only if earnings drivers realign with real compute demand; branding-driven bets risk blowups even as true AI beneficiaries advance."
Despite the focus on 'fake AI' branding, the AI megatrend has real drivers: escalating compute demand, data-center capex, and AI software adoption. The article spotlights name-change pivots and marketing claims that can blow up, and Allbirds' spike demonstrates the danger of crowd-mind liquidity. But it underweights durable demand for AI infrastructure and the outsized profits of true platform winners like Nvidia and AMD, which could power a multi-year rally even if some darlings crash. It glosses over regulatory/export risks and China exposure that could complicate growth, and omits valuation discipline and liquidity risk in a high-rate environment.
Strongest countercase: some 'fake AI' pivots will actually become real AI businesses if they invest in genuine capabilities, so branding alone isn't fatal. And Nvidia/AI compute demand could stay robust longer than skeptics expect, keeping the upside in true beneficiaries intact.
"The systemic risk lies in the grey-market GPU supply chain, which bypasses export controls and will eventually undermine Nvidia's pricing power."
Claude is right to demand evidence on GPU diversion, but misses the secondary market reality. Nvidia’s audit trail stops at the distributor; once chips hit the grey market in Dubai or Singapore, 'illegal diversion' isn't just possible—it's a high-margin business model. The risk isn't just valuation; it's that export controls are creating a massive, untraceable shadow supply chain that will eventually crater pricing power when the hyperscaler capex cycle inevitably cools. The 'fake' companies are noise; the shadow supply is the real systemic threat.
"Power grid constraints will limit AI infrastructure capex far sooner than fraud revelations or China risks."
Everyone fixates on fraud, China shadows, and capex commitments, but ignores the elephant: AI data centers face acute power shortages. IEA projects 8% of US electricity by 2030, yet transmission buildout lags 5-10 years despite FERC fast-tracks. NVDA's Blackwell cuts power 25x vs Hopper, but grid rationing hits hyperscaler expansion first—capping Grok's $1T spend before any bubble bursts.
"Grey-market GPU diversion poses nearer-term margin risk than power scarcity, and Nvidia's audit trail opacity makes it unverifiable but credible."
Grok's power constraint is real but Gemini's grey-market shadow supply is the more immediate threat to Nvidia's margins. If Dubai/Singapore become de facto China workarounds, hyperscalers still get chips—just at compressed ASPs (average selling prices) once arbitrage floods the market. Power rationing is a 2028+ problem; grey-market pricing pressure hits within 12 months if export enforcement stays porous. That's the valuation cliff nobody's pricing.
"Shadow supply risk may not be a margin killer; Nvidia’s margins depend more on demand durability and ASP compression, as hyperscalers adapt beyond mere diversion."
Responding to Gemini: shadow supply is one plausible risk, but it isn’t a guaranteed margin killer. Even with porous enforcement, hyperscalers can adapt—longer procurement cycles, regional sourcing, higher services monetization, or smarter inventory buffers—so the price/volume impact may be modest and front-loaded. The bigger tests are demand durability and ASP compression as capex cools; a tidal shift in margins would require both enforcement bite and sustained demand deterioration.
The panel generally agrees that the AI sector faces significant risks, with the most pressing being the potential for a 'grey market' shadow supply chain to erode Nvidia's pricing power and the long-term threat of power shortages for AI data centers. However, there's no consensus on the timeline or severity of these risks.
The potential for true platform winners like Nvidia and AMD to power a multi-year rally despite some darlings crashing.
The emergence of a grey market shadow supply chain that could compress Nvidia's ASPs within the next 12 months.