Looming US Power Crunch to Impair AI Abilities, Schneider Says

Schneider Electric warns that surging electricity demand from data centers powering AI could outstrip U.S. peak generation later this decade, with insufficient capacity possible by 2028 if current reserve practices persist and a projected shortfall of as much as 175 gigawatts by 2033, risking outages and blackouts. The projection signals potential near- to medium-term upside for grid upgrade and data-center infrastructure spending, higher power-market tightness and operational risk for AI-dependent services — factors hedge funds should consider for utility, infrastructure and cloud-operator exposures.

Analysis

Market structure: A 175 GW peak shortfall by 2033 (and constraints already visible by 2028) redistributes pricing power to grid-capex winners (transmission, transformers, storage) and data-center landlords who can secure priority supply. Winners: data-center REITs (EQIX, DLR), transmission/utility capex beneficiaries (NEE, EXC, GE), battery/Li-ion chain (ALB, TSLA battery units); losers: merchant baseload generators with little contracting and energy-intensive industrials without long-term offtake. Commodities and power forwards will reprice seasonally—expect upward pressure on Henry Hub and copper/conductor prices as interconnection and transformers become bottlenecks.

Risk assessment: Tail risks include rolling blackouts during heat waves, emergency price caps/regulatory moratoria on data-center builds, or a cyberattack on grid controls; any could trigger forced deratings and litigation. Time horizons: immediate (days-weeks) for weather-driven price spikes, short-term (3–18 months) for contract re-pricing and interconnection queue churn, long-term (2028–2033) for capacity deficits to materialize. Hidden dependencies: permitting/interconnection throughput, transformer lead times (6–24 months), and corporate PPA economics; catalysts include FERC/State capex approvals, IRA storage incentives, and corporate decarbonization pledges.

Trade implications: Favor long exposure to EQIX/DLR (pricing power, pass-through leases) and regulated utility capex names NEE/EXC/DUK for 12–36 month holds; hedge with short exposure to uncontracted merchant generators (e.g., short AMR/NRG-sized merchant footprints) or high-leverage, energy-intensive industrials. Take commodity plays: 3–5% portfolio exposure to front-month Henry Hub futures or short-duration nat‑gas ETFs (reduce if HH < $3/MMBtu), and 1–2% in copper futures or FCX. Use 9–18 month call spreads on EQIX/DLR and long-dated LEAPS on ALB/TSLA battery units to capture structural upside while limiting premium spend.

Contrarian angles: Consensus assumes irreversible brownouts; that understates deployment of distributed on-site generation + storage and corporate demand-response programs that can shave peaks and depress spot prices. Historical parallel: 2000s California crisis spurred both aggressive regulation and massive investment that ultimately flattened merchant margins—risk of an overbuild if incentives accelerate, creating mid-decade mean reversion. Therefore size positions modestly, watch interconnection queue throughput and FERC/state rulings for regime shifts that can rapidly flip winners to losers.

Key Decisions for Investors

• Establish a 2–3% portfolio long in data-center REITs: EQIX (Equinix) 1.5% and DLR (Digital Realty) 1.0% via 9–12 month call spreads (buy ATM call, sell 1.2x strike) to capture pricing power from constrained grid capacity; trim if power-commodity volatility compresses implied vols by >30%.
• Add 2–4% overweight in regulated/utility capex names: NEE 1.5%, EXC 1.0%, DUK 0.5% for a 12–36 month horizon; exit or hedge if regulatory filings for incremental capex are rejected or if credit spreads widen >75bps.
• Allocate 3% to short-duration natural gas exposure (front-month futures or UNG) to play near-term price shocks; set a stop-loss / reduce position if Henry Hub falls below $3.00/MMBtu for more than 30 days or if NYMEX 12-month strip declines >20%.
• Initiate a 1–2% thematic long in battery/lithium exposure: ALB (Albemarle) or TSLA battery unit exposure via 18–24 month LEAPS, sizing to 1–2% and take profits if lithium price rallies >50% or if transformer/component lead times shorten materially.
• Deploy a pair trade: long GE (grid equipment) 1% vs short NRG (NRG) 1% (or equivalent merchant generator) to express capex winners vs merchant-fired generation risk; rebalance if utility regulated ROE approvals exceed assumptions or if merchant gens secure long-term PPAs covering >60% of capacity.