Goldman: AI will save the economy someday. First, it has to stop inflating it

AI buildout is already adding inflationary pressure, with Goldman Sachs estimating about 0.3 percentage points to annual core PCE and 0.1 points to core CPI over the past year, plus another similar bump over the next 12 months. The article argues AI is pushing up costs through hardware, software subscriptions, and electricity, while evidence of productivity gains remains limited. Goldman also warned PCE inflation could reach 3.9% by May, complicating the case for near-term Fed easing.

Analysis

The market is underpricing the gap between capex-led AI adoption and monetization. In the near term, the winners are not the model vendors but the bottlenecks: memory, storage, power, and enterprise software vendors with enough pricing power to repackage AI as a feature upgrade. That favors the infrastructure layer, while consumer-facing hardware and subscription software names face a more mixed setup because price increases can support revenue but also accelerate churn and invite lower-end substitution.

The more important second-order effect is macro, not micro: AI spend is behaving like a tax on real disposable income before any productivity offset arrives. If electricity and software pricing keep feeding core services inflation, the Fed is forced to stay tighter for longer even if headline growth looks fine, which is bearish for long-duration assets and high-multiple software. The market should treat this as a duration unwind catalyst over the next 1-2 quarters, especially if the “AI productivity” narrative keeps failing to show up in aggregate data.

The contrarian point is that the inflationary impulse may be front-loaded and self-limiting. If enterprises start seeing labor substitution, procurement discipline, or reduced SaaS seat counts by late 2026, the current pricing power could reverse faster than consensus expects, creating a sharp multiple reset in the very names that have been able to raise prices on AI branding alone. That means the best short candidates are the vendors relying on AI as justification for elastic demand rather than indispensable workflow integration.

Key Decisions for Investors

• Short MSFT/ADBE/INTU basket vs long a utilities or power-generation hedge for 1-2 quarters; these names have near-term pricing power but the risk/reward is asymmetric if churn rises and AI premium pricing normalizes.
• Buy 3-6 month put spreads on AAPL; the thesis is not demand collapse, but margin compression from input-cost pass-through limits and weaker mix if non-Apple hardware prices reset upward.
• Long GS relative to mega-cap software via GS/MSFT pair for 1 quarter; GS benefits from the inflation/anti-Fed narrative and capital-markets activity while MSFT faces rising scrutiny over AI monetization quality.
• Consider a long XLU / short IGV pair into the next CPI prints; if electricity and SaaS inflation keep leaking into core services, defensive cash-flow names should outperform high-duration software by 5-10% on a relative basis.
• Add a tactical long on power infrastructure beneficiaries with liquid options only after confirmation of another hot electricity print; the catalyst window is 1-3 months, but the risk is policy intervention or load growth disappointment.