Elon Musk warns the U.S. could soon be producing more chips than we can turn on. And China doesn’t have the same issue

Elon Musk warned in Davos that U.S. AI rollout is being constrained less by chip supply than by electrical power availability, saying chips may soon outpace the ability to power them and citing data centers that could sit idle for years. China’s operational solar capacity dwarfs the U.S. (about 1,118,442 MWac vs. 237,947 MWac), while U.S. policy choices — including steep tariffs on some Asian solar imports (as high as 3,500% for Cambodia) and political resistance to renewables — are slowing deployment. Calls for PJM to boost supply, proposals for 15-year power-plant contracts to shift costs to data-center operators, and encouragement to build private nuclear plants create potential investment opportunities and risks across utilities, energy infrastructure, data-center operators and AI hardware suppliers.

Analysis

Market structure: Power constraints create a transfer of pricing power from chipmakers (NVDA) to on‑site and merchant power providers and grid upgrade contractors. Expect demand for incremental MWh to rise by low tens of TWh over 1–3 years for hyperscale AI buildouts—favors generators (NRG), utilities with build capacity (NEE, DUK) and EPC/grid equipment suppliers; U.S. solar installers and anything exposed to higher panel tariffs (residential/commercial installers) face margin compression.

Risk assessment: Tail risks include a coordinated U.S. policy shift (rapid permitting for small modular reactors or emergency capacity auctions) or prolonged permitting/legal fights that delay plant builds by 2–5 years; both would re-rate different cohorts. Near term (days–months) headline risk around PJM/FERC announcements can move generators’ stocks 10–20%; long term (2–5 years) structural underinvestment could sustain higher power prices and increase corporates’ capital intensity.

Trade implications: Direct plays are long merchant generation/peakers (NRG 6–18 months) and utility capex beneficiaries (NEE, DUK) while trimming exposure to NVDA growth beta until deployment risk is priced (reduce NVDA exposure ~10–25% near term). Use relative trades: long NEE / short RUN (Sunrun) to capture utility upside versus solar installer pain; buy protective puts on NVDA sized to 25–50% of holdings to cap a >20% drawdown.

Contrarian angles: Consensus assumes NVDA demand will convert immediately into revenue; that’s underestimating enterprise-level contracting friction and two‑year lead times on power. If PJM auctions force data centers to internalize costs, large AI buyers (MSFT, AMZN) may accelerate on‑site generation and long‑term power purchase agreements—this would benefit equipment and storage suppliers over chip multiples. A mispriced outcome is utilities with modernization plans trading cheap relative to AI multiples; look for 30–50% asymmetry in multi‑year returns.

Key Decisions for Investors

• Establish a 2–3% portfolio long position in NRG (NRG) over 6–18 months to capture merchant generation contracts with data centers; target +20–35% upside, set a 12% stop‑loss, and scale up to 4% if a PJM emergency auction is announced within 90 days.
• Reduce NVDA (NVDA) exposure by 10–25% over the next 30 days and purchase 3‑month 10% OTM puts sized to 30% of the remaining position to hedge against a >20% drawdown that could occur if deployment bottlenecks persist.
• Implement a pair trade: long NextEra Energy (NEE) 2% vs short Sunrun (RUN) 1.5% for 6–12 months to play grid/utility capex winners against U.S. solar installers hurt by tariffs; reassess after FERC/PJM decisions (monitor within 30–90 days).
• Buy call spreads on utility/infrastructure ETFs (e.g., XLU or specific names like DUK) with 9–12 month expiries to capture gradual rerating from grid investment—limit premium to 1–2% of portfolio and take profits on a 25–30% move.
• Monitor three specific catalysts over the next 30–90 days before scaling: PJM auction announcement and timeline, any emergency federal permitting for private nuclear approvals, and imposition/lifting of solar import tariffs; if any indicate faster capacity build, reallocate 50–100bp from short solar exposure into utility/infrastructure longs.