UBS: AI Stock Valuation Logic May Revert to Normal Under High Interest Rates
nashnova research
UBS's HOLT team warns that AI stocks are priced as if exempt from normal competition and financial gravity — but rising rates are testing that premise. If profitability mean-reverts, valuations would hit the historical 90th percentile, the most expensive tier.
What exactly makes AI stocks so expensive?
UBS HOLT head Michel Lerner uses CFROI — cash flow return on investment, a measure of how much cash a company earns per dollar of capital deployed — as his core metric.
His finding: if AI stocks' CFROI mean-reverts (i.e. returns fall from extreme highs back toward the long-run average), valuations would reach the historical 90th percentile — the priciest band on record.
This means → current share prices already assume these companies can earn excess profits indefinitely. If profits normalise, that support erodes.
Can semiconductors sustain their extraordinary returns?
The report projects semiconductor CFROI will triple by 2027 relative to 2024 levels.
But history is sobering: since 1950, only 1% of companies achieved that kind of explosive growth, and just 16% of those maintained equivalent CFROI five years later.
In plain terms = the number of companies that exploded in profitability *and then held on* is vanishingly small. The semiconductor sector is betting on a path almost no one has walked before.
What happens when capital spending surges?
Lerner estimates hyperscaler capex — spending by the likes of Amazon AWS, Microsoft Azure, and Google Cloud — will reach roughly $1.28 trillion between 2025 and 2028, tripling from current levels.
He cites a historical pattern: "Such investment surges have historically diluted CFROI permanently at high-return firms."
This means → the more these giants spend, the lower the cash return on each dollar invested. Scale expansion is itself the natural enemy of high returns.
What old vulnerabilities is rate-hiking exposing?
Financing risks ignored during the ultra-low-rate era are resurfacing. Lerner points to CoreWeave: as rates rose, markets began punishing deteriorating balance sheets more harshly.
Additional pressures include Chinese competitors willing to sacrifice margins for growth, and AI's disruptive impact on the software and services sector.
This reflects something broader — higher rates are not just raising the discount rate; they act as a magnifying glass, exposing every fragile spot that low rates had concealed.
Where does UBS suggest diversifying?
UK equities — tech accounts for less than 1% of the UK market's overall CFROI, making it nearly immune to an AI valuation correction.
Value stocks with momentum characteristics — particularly in financials and industrials.
Low-volatility, high-quality defensive sectors — healthcare and consumer staples, where valuations remain attractive.
Is anyone pushing back on this view?
Laura Cooper, global investment strategist at Nuveen (which manages $2.6 trillion), said Monday that "the AI theme remains intact."
Her evidence: hyperscaler order backlogs point to strong forward revenue visibility, and credit markets have yet to flash warning signals.
In plain terms = UBS sees "history says these profits won't last"; Nuveen sees "the orders are still growing right now." The divergence will ultimately be settled by whether hyperscaler capex guidance keeps delivering.
市场有风险，内容仅供研究参考，不构成投资建议。