Goldman's Wheeler Sees 'Generational Opportunity' in AI

Goldman Sachs' Miriam Wheeler says the AI infrastructure buildout is a "generational opportunity," with power demand for data centers expected to double over the next few years. The comments point to sustained investor focus on hard assets tied to AI expansion, particularly power and infrastructure capacity. The piece is commentary rather than a market-moving policy or earnings update, so near-term market impact looks limited.

Analysis

The market is starting to price AI not as a software capex story but as a utility and balance-sheet cycle. That matters because the first durable winners are likely to be lenders, infrastructure equity, and equipment suppliers with contracted cash flows, while the losers are the marginal operators with short-duration funding or commodity-like exposure to low-quality projects. In credit, this shifts relative value toward paper backed by power, grid, and data-center collateral and away from unsecured AI-adjacent borrowers whose revenues may lag the capex wave by 12-24 months.

The second-order effect is that power scarcity becomes the binding constraint, not model demand. That creates a staggered winners list: grid equipment, transformers, switchgear, gas-fired generation, and water/cooling infrastructure should outperform pure-play compute beneficiaries over the next 6-18 months because they sit on the critical path to capacity delivery. If power prices and interconnection delays stay elevated, lease-up timelines extend and leverage metrics deteriorate for speculative data-center developers, which is where the hidden credit risk sits.

The key contrarian point is that the trade is not simply "long AI." Consensus may be underestimating how much of this opportunity accrues to private and public credit rather than high-multiple equities. If capital floods into the theme, spreads on infrastructure-linked issuance can tighten quickly, but the risk is that the market overbuilds capacity before utilization catches up, compressing returns on marginal projects in 2-3 years and exposing late-cycle entrants. For now, the best risk/reward is in owning the bottlenecks, not the logos attached to the demand narrative.

GS is a modest sentiment beneficiary because leverage finance desks earn on issuance volume and sponsor activity, but the bigger impact is on underwriting mix and pipeline quality. If deal flow migrates toward asset-backed and project-finance structures, the bank can see better fee durability with lower cyclicality, though that also increases exposure to execution and refinancing risk if rates stay restrictive.

Key Decisions for Investors

• Go long XLU vs. short a basket of high-multiple AI software names over 3-6 months: infrastructure should monetize faster than application-layer AI, with lower downside if the narrative de-rates.
• Buy GS on pullbacks for a 6-12 month hold: leverage-finance and infrastructure-capex issuance should support fee pools, but size modestly because the upside is more about steady earnings durability than multiple expansion.
• Long utilities/grid-equipment beneficiaries such as VRT, ETN, and PWR into any 5-10% sector pullback: these are the cleanest bottlenecks in the AI power chain, with asymmetric upside if data-center buildouts keep accelerating.
• Prefer investment-grade or asset-backed credit tied to infrastructure over unsecured AI-linked names for the next 12 months: the spread compression opportunity is better in hard-asset cash flows than in venture-style growth stories.
• Use call spreads on a data-center infrastructure basket for 6-12 months rather than outright equity: the theme is strong, but interconnection delays and permitting can create volatility; spreads cap premium while retaining convexity.