UBS Warns AI Stocks Must Now Play by Normal Valuation Rules

Tech now accounts for roughly half of U.S. stock market capitalization, and UBS strategist Michel Lerner is asking the question that follows: "who is the next marginal buyer?"

Lerner heads HOLT, an independent value-based research product within UBS. In a note published Tuesday, he warned that AI-fueled stocks have re-rated to levels that no longer make sense in a higher-yield world, and he recommended diversifying beyond tech.

"Valuations have re-rated to imply that AI stocks are exempt from normal competitive market dynamics and basic corporate finance principles," Lerner said.

Why do valuations suddenly matter again?

The short answer: interest rates. During the era of ultra-low rates and financial repression that followed the global financial crisis, markets could shrug off deteriorating balance sheets. With rates much higher now, that tolerance is fading, and expensive valuations have started to matter. That's why AI names are beginning to underperform, Lerner argues.

The numbers behind his caution are stark. Cash flow return on investment for AI stocks sits at record levels. If that metric were to mean revert, valuations would rise to the 90th percentile of most expensive, according to his analysis.

The semiconductor picture is even more stretched. CFROI for semiconductor stocks is on target to triple from 2024 levels by 2027. Since 1950, only 1% of companies have managed growth that explosive — and of those, only 16% still enjoyed the same CFROI five years afterwards.

What about the capex boom?

Capital expenditure is a second fault line. Between 2025 and 2028, Lerner expects hyperscaler capex to triple to about $1.28 trillion. Historically, he wrote, "these investment surges have permanently diluted the CFROI of high-return businesses."

The capex boom has also created a funding risk that markets largely ignored when money was cheap. He points to the travails of CoreWeave as an example of what happens when investors start penalizing leveraged balance sheets again.

Lerner cites two further threats to the sector: cheaper Chinese competition that prioritizes growth over margin, and the existential menace AI itself poses to the software and services sector.

Where does UBS suggest hiding out?

To protect against a lasting AI downturn, Lerner backs three baskets of assets:

• U.K. equities, where tech contributes less than 1% of the aggregate CFROI
• Value stocks with momentum, especially financials and industrials
• Low-volatility, high-quality names with positive momentum and attractive valuations, such as healthcare and consumer staples

The U.K. market is his most contrarian call. It offers exposure to an economy where the AI trade barely registers in aggregate cash flows, insulating investors from a sector-specific shock.

Is anyone arguing the other side?

Yes. Laura Cooper, global investment strategist at $2.6 trillion asset manager Nuveen, said in her global macro musings published Monday that "the AI theme is intact."

For Cooper, the AI buildout isn't over — it's just bumpier. She remains constructive on the AI cycle while acknowledging that elevated expectations leave less room for disappointment. She argues the backlog for hyperscalers signals strong future revenue visibility, and that credit markets are not yet a canary in the coal mine.

The divide between the two camps comes down to one variable: whether higher rates force AI's record cash returns back toward their historical mean. If they do, UBS's math puts the sector among the most expensive deciles in market history — and the marginal buyer question gets answered the hard way.