‘Inflationary surge’: Fed economists warn AI hype is overheating the economy whether or not the technology ever delivers

TFP growth has averaged 1.11% annually since ChatGPT's 2022 launch versus a 1.23% historical average, and St. Louis Fed economists warn that AI-related 'news shocks' could spur a short-term inflationary surge as households and firms increase spending and investment on expected future gains. CPI rose 0.3% month-over-month and 2.4% year-over-year, and the authors say realized productivity gains would lower inflation over time, but if gains fail to materialize the economy risks prolonged weak growth with persistently elevated inflation. Structural demand for AI infrastructure is strong — data center vacancy at ~1.4% and about $700 billion of concentrated AI infrastructure investment — but the timing and magnitude of productivity payoffs remain highly uncertain.

Analysis

The current investment surge into specialized compute and software is creating highly concentrated demand for a narrow set of inputs (GPUs, bespoke power/cooling, specialized real estate and wafer fab capacity). That concentration is amplifying margin dispersion across incumbents: large cloud/hyperscalers can internalize capacity and capture scarcity rents while mid‑tier software and services firms face higher input costs and longer lead times, widening cross‑sector ROIC differentials over the next 6–24 months.

Because realization of productivity gains is both path‑dependent and lumpy, the economy faces a regime choice: fast, broad diffusion reduces unit costs and reintroduces disinflationary slack; slow or localized gains lock in higher prices and misallocated capex. Key dynamic risk is endogenous — if firms scale prematurely into expensive, low‑utilization capacity, we get an earnings cycle of write‑downs and capex pullbacks that can compress multiples in a single reporting season.

Second‑order beneficiaries are capital providers to the buildout (equipment OEMs, utilities serving hyperscale campuses, and specialized industrial landlords) rather than pure software vendors; losers are service chains exposed to cyclical CRE services and wage stickiness in high‑skill labor markets. Watch cross‑market signals — cloud utilization, power reserve margins, equipment lead times, and capex guidance — as the fastest indicators of whether productivity is diffusing or investment is overheating.

Near term (weeks–quarters) the tradeable story is dispersion and optionality: favor assets that monetize scarcity and offer contracted cash flows over those priced for broad productivity upside. Over 12–36 months, monitor adoption metrics — if productivity diffusion accelerates, re‑rate software and broad service plays; if not, expect a protracted re‑pricing toward capital owners of infrastructure.

Key Decisions for Investors

• Long Digital Realty (DLR) — 12–18 month horizon. Rationale: contracted cash flows tied to hyperscaler demand and ability to pass-through higher power/colocation pricing. Target +25% upside; downside risk ~-30% if hyperscalers pivot to in-house capacity. Position size: 3–5% sector tilt with 15% stop-loss.
• Long Lam Research (LRCX) or ASML (ASML) — 6–12 month horizon. Rationale: equipment OEMs benefit earlier in the build cycle and have more durable orderbooks; play semiconductor capex cadence. Target +30% upside; downside ~-35% on macro pullback. Use 9–12 month call spreads to cap premium cost.
• Pair trade: Long Equinix (EQIX) / Short CBRE (CBRE) — 9–12 month horizon. Rationale: long pure‑play data center REIT with recurring revenue vs short cyclical CRE services provider exposed to volatile leasing markets. Aim for 2:1 reward:risk (target net +20% relative return; max drawdown on the pair -15%).
• Tactical options hedge: Buy 12‑month broad market put spread (e.g., SPX 5–7% wide) for 0.5–1% portfolio cost to protect against a sudden capex‑reset shock that derates cyclicals. Reduce cost by selling a further OTM put.
• Trigger monitoring & exit rules: trim infrastructure longs if 1) cloud utilization drops meaningfully in two consecutive weeks, 2) equipment lead times shorten by >25%, or 3) top‑line guidance from two large hyperscalers signals capex pause; escalate to full review if any occurs.