What will power the grid in 2035? The race is wide open

40% of U.S. natural gas consumption currently goes to power generation, and tech-driven AI demand has accelerated investment in SMR and fusion startups that target grid connection in roughly 5–7 years. Cost and scale are major constraints: new nuclear ≈ $170/MWh, early fusion estimates ≈ $150/MWh, new baseload gas ≈ $107/MWh, while solar+storage ranges $50–$130/MWh and battery deployments reached 58 GWh last year. Ambitious company targets (e.g., Helion aiming for 5 GW by 2030 and 50 GW by 2035) would require building ~800 reactors by 2030 and ~7,200 more by 2035, highlighting substantial execution and capital risk despite meaningful long-term disruption potential to the gas market.

Analysis

Tech-sector pre-commitments to next-generation baseload (via equity, PPAs, and JV terms) are creating a multi-decade procurement advantage that is invisible on balance sheets today: startups that secure anchor customers will be able to compress cost-of-capital, de-risk serial manufacturing runs, and erect effective moats through long-term offtake and site learning curves. That dynamic favors platform-scale cloud vendors and hyperscalers that can underwrite multi-gigawatt projects as a form of strategic R&D and supply diversification, while imposing a two-sided squeeze on incumbent merchant generators (who face demand erosion) and capital goods suppliers (who may see concentrated orders from a few large new entrants).

The principal tail risks are execution and unit-economics: serial fabrication must drive >30-50% CAPEX declines versus bespoke builds to make SMR/fusion competitive with mature combined cycle economics, and regulatory/interconnection cycles remain an underpriced timing risk that can stretch outcomes from quarters to decades. Shorter-horizon catalysts to watch are first-grid demonstrations showing sustained capacity factor, dispatchability over multi-day events, and published LCOE below $100/MWh; a single credible commercial contract that delivers sub-$100/MWh for multiple years would reprice expected demand curves for thermal gas. Geopolitical shocks or breakthrough long-duration storage (multi-day iron-air / flow chemistries) are asymmetric events that can rapidly accelerate either the incumbent or new-technology pathways.

From a competitive-structure view, a bifurcation is probable: firms that master modular supply chains and standardize site integration will capture outsized margins, while standalone prototype vendors risk consolidation or venture-style write-downs. Expect heavy forgers, modular shipyards, and EPC firms that adapt to factory-built modules to rerate; conversely, geographies with fast permitting and grid access will become premium markets for deployment, concentrating value rather than distributing it widely.

Key Decisions for Investors

• Buy a small, staged equity stake in OKLO (size 1-2% NAV) as a binary asymmetric bet: entry now, add on demonstrable construction milestones or attractive PPA announcements. Reward: >3x if they scale serial builds; Risk: high dilution/regulatory set-backs — hard stop at 40% drawdown or failure to meet a key licensing milestone within 18 months.
• Buy long-dated call LEAPS on MSFT (12–36 month expiries) to capture the margin upside from lower and more predictable power procurement costs for its data center fleet and continued AI-driven revenue growth. Position sizing 2-3% NAV; target 2.5x upside if cloud monetization and offtake contracts improve margins; downside limited to premium paid.
• Buy GOOGL (or GOOGL Jan-2029 LEAPS) selectively as a hedge-rich exposure to a company able to finance and lock long-term clean baseload; expect reduced energy cost volatility to incrementally expand AI gross margins over 2–5 years. Size 2% NAV, take profits on 30–50% outsized re-rating tied to disclosed PPAs/ownership stakes.
• Hedge the thematic risk by a small short of natural gas price exposure (UNG or short-term natural gas producers, size 1% NAV): rationale is that successful scale-up of long-duration storage or modular nuclear/fusion would structurally lower baseload demand. Use tight stops (15–25%) because supply shocks can temporarily invert the thesis; unwind if 3-month realized volatility in nat gas > 60%.