Goldman Sachs favors hyperscalers over chipmakers in AI infrastructure By Investing.com

Goldman Sachs’ Jim Covello said investors should favor hyperscalers over chipmakers in the AI infrastructure buildout, arguing the market has already priced in skepticism on hyperscaler ROI. The note highlights that semiconductors have surged nearly 150% over the past year while hyperscalers such as Amazon, Oracle, Microsoft, Alphabet and Meta have lagged as capital-spending concerns persist. The call is a relative-value rotation view rather than a broad market macro catalyst.

Analysis

This is less a fundamental call than a positioning signal: the market has already crowded into the “picks-and-shovels” AI trade, so the next leg is more likely to come from beneficiaries whose spending is being doubted rather than the suppliers already marked for perfection. If hyperscalers can show even modest evidence that incremental AI capex is translating into attach rates, monetization, or lower unit costs, the rerating could be larger than the earnings delta because their valuation compression has created optionality on sentiment recovery. The key second-order effect is that a shift in leadership would force systematic reallocations out of semis and into mega-cap software/platform exposures, amplifying the move through passive and factor flows.

The risk is timing: chip earnings may remain structurally stronger for 1-2 quarters even if the trade is crowded, while hyperscaler ROI proof likely takes multiple quarters and may not show up cleanly in reported numbers. That creates a squeeze risk for anyone short the semiconductor basket too early, especially given how much momentum remains embedded in AI infrastructure beneficiaries. The cleaner reversal trigger is not “good AI demand,” but evidence that capex intensity is flattening while monetization per dollar spent is rising — a setup that could happen over the next 6-12 months if cloud AI usage starts to scale faster than depreciation growth.

The contrarian angle is that investors may be underestimating how much of the hyperscaler underperformance is already a discount for capex overhang, meaning bad news could be largely in the price while chip multiples still have less room to expand from here. If the AI buildout enters a digestion phase, suppliers tied to near-term orders can disappoint on growth rates even as the end-user platforms improve relative returns. In that regime, the best trade is not to bet on absolute AI weakness, but on the rotation from “spend enablers” to “monetizers.”

Key Decisions for Investors

• Initiate a 3-6 month pair trade: long AMZN/MSFT/GOOGL basket vs short SMH or SOXX. Risk/reward favors the long side if capex-to-revenue conversion improves; stop if semiconductor relative strength continues to expand for another earnings cycle.
• Buy 6-12 month call spreads on MSFT or AMZN rather than outright stock to express a rerating on capex ROI. Best setup is into any post-earnings dip where investors punish spending without evidence of monetization deterioration.
• Reduce outright long exposure to semis into strength, especially names most exposed to AI capex beta, and rotate toward hyperscalers with the highest operating leverage to AI usage. The risk is missing one more momentum leg, so scale out rather than liquidate.
• For more tactical exposure, consider a short-dated downside hedge on a semiconductor ETF against existing AI winners. The thesis breaks only if hyperscaler commentary remains weak on returns while chip order flow stays exceptionally strong.