Goldman Sachs Strategist: Core AI Stocks Are Undervalued, Recommends Adding Exposure Across Three Key Themes
Alina Collins
Goldman Sachs chief U.S. equity strategist Ben Snider says AI stock valuations signal opportunity, not a bubble, and names AI infrastructure, power infrastructure, and hyperscale cloud as three themes worth adding — arguing that lingering market skepticism is itself a margin of safety.
Are AI stocks too expensive — and why does Goldman say "no"?
Two opposing fears coexist: some investors worry AI stocks are overvalued, while others are uneasy that valuations look low relative to earnings growth. Snider sees the split itself as proof that the market has not gone all-in.
This means → the equity risk premium — the extra return investors demand for owning stocks over bonds — has not been fully compressed. There is still skepticism priced in.
In plain terms = bubbles form when everyone has already piled in. Widespread hesitation is what makes this a buying window, not a warning sign.
The Shiller P/E is near record highs — why isn't Snider worried?
The Shiller CAPE ratio — a price-to-earnings measure smoothed over ten years of average earnings — sits close to its all-time high, on par with the dot-com peak and the 2021 top.
Yet the S&P 500 has gained over 20% in the past 12 months while its 12-month forward P/E is actually lower than a year ago.
This means → the rally is driven by earnings growth, not by investors paying a higher multiple for the same profits. This reflects a healthier quality of gains than a pure valuation expansion.
Theme one: AI infrastructure — why does Snider call it undervalued?
Covers semiconductors, servers, and AI networking equipment, including memory-chip makers.
Snider notes that valuation multiples across the semiconductor sector have seen almost no expansion, signaling persistent market doubt about future earnings.
In plain terms = chip companies' profits are already rising, but their stock prices have not kept pace — the market "doesn't believe it yet," and that gap is the discount.
Theme two: power infrastructure — AI's utility bill?
Snider calls this theme "extremely attractive." Two drivers: AI's relentlessly growing electricity demand, and the impact of the Iran conflict on energy infrastructure.
Power infrastructure has become a core pillar of the 2026 "HALO trade."
This means → AI is not just a software story. Whoever can deliver power to data centers sits on a bottleneck the entire supply chain must pass through.
Theme three: hyperscale cloud — a dip worth buying?
Includes Amazon, Microsoft, Meta, Alphabet, Oracle, and IBM. All have outperformed the broad market since late 2022, but a sector rotation this year has hit them hard — they "just posted one of their worst months on record."
On a P/E basis, they now trade at the bottom of their ten-year range, near levels last seen in March 2020 or late-2022 bear-market lows.
In plain terms = the fundamentals have not deteriorated; money simply rotated elsewhere, compressing valuations to multi-year lows — exactly the "add exposure" window Snider describes.
The biggest risk — could cloud giants suddenly slash AI spending?
Snider views this probability as low: a sudden cut in AI capex by hyperscalers is unlikely, and sustained growth in AI buildout spending remains the core catalyst for the entire chain.
That said, this call will face its test in the upcoming earnings season.
This means → next quarter's results are the key verification window. If hyperscaler capex guidance holds, the three-theme thesis stays intact; if it doesn't, a reassessment is warranted.
Content is for reference only, not financial advice.