MW A research firm's 'obituary' for software stocks vaporized billions. Here's the part the market is finally admitting.
By Charlie Garcia
The Citrini selloff wasn't panic - it was an X-ray. The 'smart money' is quietly exiting the tech sectors AI is disrupting.
A recent bombshell report on AI tech tanked shares of vulnerable software and financial services companies.
Citrini Research's report detailed ways that AI tech could dominate and disrupt. You still have time to reposition your portfolio.
Citrini Research, the No. 1 finance newsletter on Substack with close to 200,000 subscribers, published a scenario analysis about artificial intelligence. It's a scenario, not a prediction.
"The 2028 Global Intelligence Crisis" published last month imagines AI unleashing both a productivity boom and widespread unemployment. Cue the AI scare: Michael Burry of "Big Short" fame posts on X: "And you think I'm bearish." Commentator Nassim Taleb amplifies the volume with bankruptcy warnings.
The Citrini report posted on Sunday, Feb. 22. By the next day's close of trading, the Dow had shed 821 points. IBM suffered its worst day in 25 years, compounded by Anthropic's announcement that its Claude AI could modernize COBOL, the programming language holding IBM's mainframe empire together. Software stocks dropped 4%. The banking sector posted its worst session since April 2025.
Here's the part nobody wants to talk about: The Citrini piece didn't move markets because it was scary. It moved markets because the people trading those stocks recognized their job descriptions in the obituary section.
That's not a market reaction. That's an X-ray.
Which assumptions won't survive?
I profiled Citrini founder James van Geelen for MarketWatch in January. He called Nvidia (NVDA) early. He called GLP-1s early. He shorted Silicon Valley Bank before it collapsed and turned Secured Overnight Financing Rate $(SOFR)$ call options into a 46x return.
Van Geelen named his firm after citrinitas, the alchemical stage where lead starts becoming gold but isn't gold yet. The inspiration came from George Soros's "Alchemy of Finance," the first book that made van Geelen see investing as philosophy rather than spreadsheets. He lives in the transition before the transformation completes. That's where his money is made.
His AI thesis isn't complicated. For all of economic history, human intelligence was the scarce input. Capital was abundant. Resources were substitutable. But the ability to analyze, decide, coordinate and bill $400 an hour for it? That was the franchise.
AI is about to commoditize the franchise.
Start with white-collar employment. Information-worker jobs are already down roughly 8% from their 2023 peak. Big Tech has been executing round after round of layoffs while reporting record margins. The displacement isn't a forecast. It's a line item.
There's also SaaS (software as a service) pricing power. Van Geelen tells a story about OpenAI's forward-deployed engineers being used as procurement leverage to extract 30% discounts from software vendors. When your customer's AI vendor helps it negotiate against you, your pricing power isn't eroding. It's being actively demolished.
Then the big one: intermediation. Not just software. All intermediation. Insurance companies intermediating risk. Banks running credit cards, lending and deposits. Financial advisers charging 1% for what a competent AI portfolio tool replicates for close to free.
The moat for all these businesses isn't technology. It's friction. Friction is precisely what agentic AI eliminates. Imagine it's two years from today and most of what you do on your phone, your AI agent instead does for you. You're a lot less tied to a website.
Here are your 21st-century whip makers
Van Geelen's test is ruthless: If an AI agent can do it for you, a company's competitive edge was never its technology. It was your laziness.
If the moat was friction, not technology, the following sectors are exposed.
Long-tail SaaS companies whose entire value proposition is workflow automation that a foundation model does better and cheaper. Vulnerable companies include HubSpot (HUBS), Asana $(ASAN)$, Monday.com $(MNDY)$ and Atlassian $(TEAM)$. Wall Street has already started calling this the "SaaSpocalypse."
Add payment processors collecting interchange on every swipe: Visa (V), Mastercard $(MA)$, American Express $(AXP)$. When AI agents can circumvent the rails and stablecoins settle on-chain, this toll booth disappears.
Also vulnerable are financial advisory firms charging percentage-of-assets fees in a world where AI agents comparison-shop across vendors while you sleep, companies such as LPL Financial Holdings $(LPLA)$, Charles Schwab $(SCHW)$, Raymond James Financial $(RJF)$ and Ameriprise Financial $(AMP)$. A fintech startup called Altruist recently launched an AI tax-planning tool for $60 a month, and LPL shares lost 8% in a day.
Van Geelen's test is ruthless: If an AI agent can do it for you, a company's competitive edge was never its technology. It was your laziness.
But investors should understand something. That Citrini report wasn't neutral analysis. Co-author Alap Shah, CIO of Lotus Technology Management, confirmed on Bloomberg TV that his firm was short names in these sectors. Shah is also CEO of Littlebird, an AI agent company. If agents disrupt the economy as described, his company benefits. Everybody at the table was playing with real chips.
This isn't an accusation. This is smart-money positioning you should pay attention to.
One more thing: Apollo Global Management $(APO)$, the biggest private-credit player in the room, quietly reduced its software lending in early 2025. When the smartest leveraged money on the planet exits a sector before anyone's talking about it, that is the signal. You don't need a Substack essay. You need Apollo's loan book.
The winners
Van Geelen's winners list tracks closely with what I've been telling readers for years. Which is satisfying in the way that being early to a thesis usually isn't.
1. Semiconductors: The companies making the chips that power the intelligence revolution are the new utilities.
2. Gold (GC00) and silver (SI00): Both are up dramatically from a year ago. Hard assets don't get disrupted by agentic AI. Gold doesn't get automated. Silver doesn't get a 30% procurement discount. And there's bitcoin (BTCUSD), down from highs but structurally sound in a world where the "intelligence premium" on human labor is unwinding and digital scarcity still matters.
3. Energy: Data centers need power. Oil (CL00) is now trading in the $100 vicinity. The physical economy is reminding the digital economy who signs the electricity bill.
4. Systems of record: Van Geelen's contrarian call here is sharp. The big enterprise platforms are positioned for what he describes as a short squeeze. AI slashes their coding costs while maintaining their revenue. That's margin expansion, not destruction.
5. Brands: Luxury. Identity. Status. Things AI can't replicate because their value is social, not functional. Hermès doesn't care if your AI agent found a cheaper handbag.
What you're not seeing yet - but will
Few are tracking Chinese consumer data as a leading indicator for U.S. white-collar disruption. You should be.
Van Geelen's sharpest insight didn't come from last month's Citrini piece. It came from his recent appearance on the Bloomberg Odd Lots podcast, which most investors haven't listened to yet.
China, he says, may be foreshadowing what will happen in the U.S. over the next two years as automation rippled through, jobs stopped being produced, the consumer economy weakened and everything was dragged down.
Don't miss: China's middle class is quietly stockpiling cash - 'shadow saving' - and it could short-circuit the global economy
Few are tracking Chinese consumer data as a leading indicator for U.S. white-collar disruption. You should be.
Here's the data: Chinese retail sales grew 2.1% in early 2026 - a historical low apart from the COVID pandemic. Consumer prices have fallen for six straight quarters. Youth unemployment hit 16.9%, and those same young people account for 45% of the downturn in spending. Households are the only sector deleveraging. They're not spending because they're bracing.
Automation hasn't even fully arrived. This is the preview reel.
Then there's the composition lie. Citadel's rebuttal points to software job postings up 11% on a year-over-year basis. Sounds reassuring - until you realize those postings are 70% below the 2022 peak. Up 11% from a crater is still in the crater.
Now look inside the 11% - AI, ML and cybersecurity roles alone drove the entire growth. ML engineering is machine-learning engineering - building, deploying and operating AI models. Entry-level tech postings are down 34%. The old jobs are dying. The new jobs require completely different skills. The aggregate number doesn't lie. It hides a structural transition behind a single percentage point.
Van Geelen's milestone to watch: white-collar employment - specifically information workers. Five percent losing jobs in 18 months without other white-collar jobs to go to. They flood blue-collar and gig markets, depressing wages everywhere. It's all one closed system.
And the reaction itself was the signal. When a single research report vaporizes hundreds of billions in market cap, institutional investors have already modeled their own displacement. They're one narrative away from panic. That's a fragility indicator you can trade around.
The canary lives - for now
Citrini's essay ends with a line that deserves more attention than the doom that preceded it: "The canary is still alive."
Which means you still have time to reposition. Audit your portfolio for friction-dependent businesses. Add exposure to hard assets, semiconductors and energy. Watch Chinese consumer data like a hawk.
And when the next bombshell research report drops, ask yourself: Do I own the companies in the obituary section or the companies writing the eulogy?
(MORE TO FOLLOW) Dow Jones Newswires
March 19, 2026 08:19 ET (12:19 GMT)
Copyright (c) 2026 Dow Jones & Company, Inc.
Comments