The AI bubble fears are getting worse. Wall Street doesn't want to talk about it

A version of this article originally appeared in Quartz’s members-only Weekend Brief newsletter. Quartz members get access to exclusive newsletters and more. Sign up here.

When even Sam Altman says investors are "overexcited," you know something's up.

The OpenAI CEO's recent warning joins a growing chorus of voices — from MIT researchers finding 95% of AI projects unprofitable to Apollo economists comparing today's valuations to the dot-com era. Yet as these concerns mount and stocks like Nvidia slip despite record earnings, much of Wall Street continues its elaborate linguistic dance around the one word that could make sense of it all: bubble.

The latest warning came from an unexpected messenger. At a small dinner with reporters last month, Altman, reportedly seeking funding at a $500 billion valuation, delivered an unexpectedly sober assessment for someone leading the AI charge.

"When bubbles happen, smart people get overexcited about a kernel of truth," he said. "Are we in a phase where investors as a whole are overexcited about AI? My opinion is yes."

It wasn't just idle speculation. GPT-5's underwhelming debut had already rattled some investors' faith in AI's inevitable march forward. Altman was joining a chorus that had been growing louder for months, not starting one.

Skeptics such as NYU's Gary Marcus have been warning about AI's limitations for years, and of a potential bubble around generative AI since 2023. Apollo Global Management's Torsten Slok has been sounding alarms about valuations since July. 

But when the face of the AI boom himself starts using bubble language, it carries different weight.

The data backing up these concerns has been mounting. MIT researchers found that 95% of corporate generative AI projects have failed to generate profit. Slok calculated that today's top tech companies are more overvalued than their dot-com predecessors. Even former Google CEO Eric Schmidt seemed to retreat from his earlier AGI optimism in a recent New York Times op-ed.

Record results, falling stock

Nvidia’s earnings last month offered a perfect case study in the market's strange psychology. The AI chipmaker posted record quarterly sales of $46.7 billion, with data center revenue rising 56% to $41.1 billion. By most measures, it was a blockbuster quarter.

The stock fell almost 3% in after-hours trading.

The disconnect wasn't lost on analysts. Despite the eye-popping revenue numbers, Nvidia had missed data center expectations for the second consecutive quarter, a concerning pattern for a segment that accounts for 89% of the company's sales. The company's $54 billion revenue guidance for the third quarter, while higher than estimates, was described as "underwhelming" and "lukewarm."

Wall Street finds itself caught between optimism and anxiety, deploying different vocabularies depending on which side of the ledger they're addressing. The same banks that tout AI's transformative potential also warn clients about potential pitfalls.

Morgan Stanley sees efficiencies worth almost a trillion dollars annually from AI adoption, while UBS speaks of "capex indigestion" when companies struggle to digest massive infrastructure spending. Bank of America frames the spending as an "innovation premium" driving a productivity "sea change," but elsewhere warns of companies being "re-rated too aggressively."

The split reflects the industry's conflicted position. These banks are simultaneously cheerleading the AI boom by financing data center buildouts, underwriting deals, and managing AI-heavy portfolios, while privately harboring concerns about sustainability. 

Speaking too plainly about bubble conditions would undermine their own business. But ignoring the risks entirely would be professional malpractice.

The stakes are high

The reluctance to name what's happening isn't just semantic squeamishness: There's too much riding on the AI buildout for anyone to want to spook the market. AI infrastructure spending has contributed more to U.S. economic growth in recent quarters than consumer spending, effectively becoming a private-sector stimulus program. The Magnificent 7 tech companies alone spent a record $102.5 billion on capital expenditures in their most recent quarters, almost all of it flowing to data centers and AI infrastructure.

It's an echo of the late 1990s, when massive spending on fiber optic cables and telecom infrastructure drove economic growth even as dot-com valuations soared beyond reason. The buildout continued right up until the crash, leaving behind valuable infrastructure that eventually powered the next generation of internet companies. But it also left behind massive overcapacity and trillions in losses.

Of course, the skeptics could be wrong. AI adoption is accelerating, and unlike many dot-com companies, today's tech giants have massive existing cash flows to support their infrastructure bets. The current AI boom is barely two years old, making definitive judgments difficult. 

Until the technology proves itself over time, Wall Street's linguistic dance around the B-word may be the most rational response — acknowledging both the promise and the peril without committing to either narrative.