AI broke the energy grid’s assumptions

AI-driven data-centre growth is creating 'massive, uninterrupted' 24/7 electricity demand and driving a multibillion-dollar allocation race across gas, nuclear fission (SMRs), fusion and advanced renewables. A key constraint is gas-turbine supply: manufacturers' lead times push deliveries into the early 2030s, creating a roughly 5–7 year window for alternatives to commercialize. SMR developers target commercial starts from 2028–2030+ (Oklo 2028, TerraPower 2030, X-energy early 2030s; Kairos has installed an ETU vessel), while fusion is moving toward a commercial timeline—any of which could materially reshape power markets and capital deployment.

Analysis

The sudden, structural uplift in always-on electricity demand changes counterparty economics: large tech buyers will trade price for certainty, locking developers into long-dated, capacity-guaranteed contracts that shift cash-flow risk away from merchant generators and toward firms that can deliver turnkey, schedule-assured output. That creates asymmetric optionality for technology-native suppliers (factory-built modular systems, niche industrial OEMs) and for buyers willing to underwrite construction risk in exchange for guaranteed baseload pricing.

A supplier-side bottleneck creates two profitable dynamics for incumbents: pricing power on scarce equipment and a multi-year window to vertically integrate high-assurance inputs (forgings, instrumentation, qualification services). Private-equity-style consolidation opportunities emerge in specialty industrials whose revenues are small relative to addressable demand; these names are under-owned by public energy funds and could rerate quickly if they secure multi-year contracts.

The biggest near-term macro swing is regional fuel-basis divergence. Competing uses for gas (exports, industry, power for new always-on loads) will widen basis and spark spreads unevenly, creating arbitrage across ISO hubs and making localized PPAs and on-site solutions economically rational even at higher per-MWh prices. Key reversal catalysts are regulatory acceleration (fast-tracked permits/licensing), a sudden surge in OEM capacity from a single new factory, or an economic slowdown that deflates data-centre expansion — each could flip winners into losers within 6–36 months.

Key Decisions for Investors

• OKLO (OKLO) — small asymmetric call-spread: buy a Jan-2029 call / sell a higher strike Jan-2029 call representing 0.25% NAV. Rationale: binary upside if licensing/construction milestones hit; capped premium loss if milestones slip. Timeframe: 18–36 months. Risk/Reward: potential 5x–10x on premium vs full loss of premium if delays/denials occur.
• GE (GE) — directional equity + tail protection: accumulate GE (1–2% NAV) to capture aftermarket pricing and spare-parts margin expansion; buy 12-month OTM puts sized to cap downside at ~50% of position. Timeframe: 9–18 months. Risk/Reward: asymmetric — modest upside if backlog monetizes, controlled downside via put protection if demand pivots to alternatives.
• Digital infrastructure pair: long Digital Realty (DLR) / short Duke Energy (DUK) 1:1 market-cap-adjusted (net exposure 1% NAV). Rationale: data-center landlords benefit from long-term demand and willingness to pay for priced certainty; regulated utilities face rate pressure and slower growth if corporates bypass grid. Timeframe: 12–36 months. Risk/Reward: captures spread between growth-forcing demand for capacity and regulated growth compression; downside if broad slowdown hits tech capex.
• Specialty industrials (ATI) — selective small-capitalization exposure: allocate 0.5% NAV to high-integrity alloys/forgings suppliers (example ticker: ATI) via equity or long-dated calls. Rationale: under-owned suppliers can reprice quickly once multi-year supply contracts crystallize. Timeframe: 12–36 months. Risk/Reward: low-to-medium probability of re-rating but outsized returns if a handful of multi-year contracts are awarded.