Goldman flags three key ways in which AI is boosting consumer prices

Goldman Sachs said AI is already adding about 0.3 percentage points to annual core PCE inflation and 0.1 percentage point to core CPI, with a further 0.1 to 0.2 percentage point lift to headline PCE possible from higher electricity prices over the next several years. The bank identified three inflationary channels: higher electronics input costs, software price increases tied to AI features, and rising power demand from data centers. Goldman expects these pressures to fade eventually as AI-driven productivity gains lower costs and make the technology disinflationary over time.

Analysis

The important second-order read-through is that AI is behaving like a late-cycle capex shock before it behaves like a productivity shock. That means the near-term macro effect is a squeeze on margins and real household spending power, while the medium-term benefit accrues unevenly to firms that can amortize AI investment over a large installed base. In practice, this is more bullish for infrastructure-heavy vendors and utilities than for software names trying to reprice seats faster than customers can accept.

The competitive dynamic is asymmetric inside software: firms with high switching costs and mission-critical workflows can push through AI surcharges, but consumer-facing or discretionary SaaS brands face a higher risk of churn or downgrade pressure once customers see AI fees as a tax rather than a feature. That makes pricing power the key discriminator, not "AI exposure" as a theme. The bigger hidden loser may be adjacent categories that depend on consumer PC refresh cycles and enterprise device upgrades, because higher input costs and tariff-like price increases can delay procurement into 2026.

On inflation, the market should not extrapolate the current AI impulse linearly. Electricity and component bottlenecks are front-loaded, but they are also the easiest areas for supply to respond if capex stays elevated; the disinflationary payoff from productivity is slower, broader, and more likely to show up in labor-intensive sectors than in tech margins. The most likely base case is a 6-18 month window where AI is mildly inflationary at the headline level, while simultaneously widening dispersion between AI-capable firms and everything else.

The contrarian point is that consensus may still be underestimating how much of this cost pressure is temporary and inventory-driven. If memory and battery pricing normalize sooner than expected, the inflation narrative can fade quickly, while the productivity story remains intact for the vendors that win the platform layer. That argues for trading the spread between firms with real pricing power and those using AI as justification for price increases without corresponding utility gains.

Key Decisions for Investors

• Long GS vs short QQQ on a 3-6 month horizon: own the bank that benefits from pricing/flow volatility while fading a basket where valuation remains vulnerable to margin pressure if AI capex keeps inflation sticky.
• Short ADBE / DUOL / INTU as a basket into the next earnings cycle if guidance confirms AI-driven price increases are meeting resistance; risk/reward is favorable because the market can punish churn and downgrade risk faster than it rewards monetization.
• Long MSFT on a 12-month horizon, but use call spreads rather than outright stock: it is best positioned to pass through AI monetization, though near-term margin optics can be noisy from infrastructure spend.
• Long utilities and power infrastructure beneficiaries on dips, especially names with regulated rate-base growth and data-center load exposure; this is a cleaner way to express the electricity-price channel than chasing broad tech beta.
• Pair long semicap equipment / memory beneficiaries against short consumer hardware-dependent names for 6-9 months; if AI demand remains strong, the supply-chain winners get paid first, while end-market pricing pressure hits later.