Scott Galloway Predicts AI Valuations Will Crater 50-70% Within 24 Months. Here’s Why.

Scott Galloway has made his most aggressive call yet on the AI trade. On the Prof G Markets segment titled AI May Not Be Worth The Cost: Here’s Why, he argued that AI companies are dramatically overvalued and predicted a major repricing within the next 24 months. “If it’s either going to be labor chaos or valuations coming down by 50, 60, 70%, I absolutely think it’s the latter,” he said.

I’ve been investing through AI cycles for as long as you could with a long-held position in NVIDIA (NASDAQ:NVDA | NVDA Price Prediction) and Alphabet (NASDAQ:GOOGL), and I want to walk through Galloway’s case carefully before explaining where I respectfully part ways with him.

The Galloway Thesis: Bundling, Layoffs, and the Chump Train

Galloway’s first complaint is structural. He argues that Elon Musk is running a “Snow White and the Seven Dwarves” bundling strategy, where SpaceX is “ridiculously hot” with $16 billion in revenues and $8 billion in operating profits, but investors who want exposure must also accept “this money furnace called XAI.” Asked whether he would sell SpaceX shares at a potential $1-2 trillion valuation, his answer was blunt: “an immediate sell right now, today. Easy. No questions whatsoever.”

His second point is about the math behind AI productivity claims. Galloway argues AI has failed to produce meaningful new revenue-generating products beyond LLMs themselves. To justify current valuations through efficiency gains alone, he estimates the economy would need 5 to 7 million layoffs across 75 million AI-vulnerable workers, representing 10% labor destruction. With the U.S. unemployment rate sitting at 4.3% as of April 1, 2026, that kind of dislocation would be a generational shock.

He also flagged geopolitical risk, predicting China could accelerate the repricing by “dumping” cheap LLMs into the U.S. market, and he eviscerated the current IPO pipeline, calling it “the last stop on the chump train” where private investors have already “squeezed as much juice out of this as we can” before offloading to “people stupider than us.” You can hear the full segment on the Prof G Markets feed.

The Evidence Galloway Can Point To

Some of the tape supports him. Jim Cramer on Mad Money flagged Quantinuum’s IPO this week, noting the company had just $30.9 million in revenues last year while opening at a valuation over $15 billion, with a pro forma net loss of $515.6 million. Marvell (NASDAQ:MRVL) jumped 40% after Jensen Huang praised it from a stage. Insider selling has been heavy, with Silver Lake entities unloading $77.6M in Dell (NYSE:DELL) stock on June 2 after a 280% return over the past year. The VIX, meanwhile, sits at 16.06, a complacency reading after a spike to 31.05 on March 27, 2026.

Where I Respectfully Disagree

Here is where I part ways with Galloway. A 50-70% haircut across AI assets within 24 months assumes the productivity story is mostly fiction. I think it is early, messy, and underpriced in places, overpriced in others. OpenAI’s CFO has said compute through the end of 2027 is sold out, and Broadcom (NASDAQ:AVGO)’s CEO has visibility booked into 2028. That is real demand from real customers writing real checks, including hyperscalers I own.

Galloway is right that the IPO market looks like a dump-and-run, and he is right that XAI-style bundling deserves skepticism. He is also right that GDP growth has cooled to 1.6% annualized as of January 1, 2026, which leaves less room for error if AI capex disappoints.

My read: expect a 30-40% drawdown somewhere in the leaders during this cycle, possibly more in the speculative tail Cramer warned about. Transformative buildouts (railroads, electricity, the internet) all featured savage interim selloffs before the winners emerged. Galloway may be early on the volatility and wrong on the magnitude. Either way, sizing positions so you can survive a Galloway-style tape without selling is the move worth making before the next VIX spike, instead of after it.