Fueling the AI Machine: The Road to AI Leadership
From patchwork electric grids to high energy availability (2025–2065)
The Problem
The Energy Footprint of AI.
Artificial Intelligence is no longer a promising technology—it is the defining industrial race of our time. Like the steam engine, electrification, and the internet before it, AI’s growth will be driven not by algorithms, but by energy.
Every new AI capability, from training the largest models to serving billions of real-time inferences, requires electricity. Data centers, training clusters, and inference infrastructure are doubling and redoubling in demand, and projections suggest AI energy use will rise from roughly 4% of U.S. electricity in 2025 to more than 20% by 2035. If unchecked, this trajectory will overwhelm today’s energy supply.
The result: rising costs, strained utilities, and a scramble to add new energy capacity. For the first time in decades, energy supply—not advanced computer chips—has become the bottleneck for Ai leadership.
The Patchwork Grid
The U.S. electrical grid is a patchwork relic of the 1980s—designed for household and light industrial loads, not high-density AI clusters or quantum labs. Blackouts and bottlenecked transmission lines reveal how unprepared we are for the load profiles AI requires.
To sustain AI growth, a complete remaking of the grid is needed: new plants, new transmission, new baseload strategies, and smarter demand-response systems.
The Competing Objective: Climate
For the past two decades, U.S. energy policy has been dominated by climate ambition. Through subsidies, tax credits, and regulatory preference, the country has funneled capital into renewables—particularly solar, wind, and geothermal.
Average subsidy rates: ~20% for renewables versus ~6% for fossil fuels over the long term.
The effect: subsidies flattened early risks, attracted private capital, and successfully built a competitive renewables industry.
That goal was achieved. But the arrival of the AI era changes the priority.
AI’s Energy Needs vs. Renewables
AI doesn’t sleep. It requires 24/7 baseload electricity, not intermittent generation. Solar and wind, even with storage, cannot yet provide the density and reliability needed for AI to scale.
Putting environmental concerns aside momentarily, the hard truth is this: renewables alone will not power AI’s future.
The key decision for the next 40 years is whether the U.S. will treat energy as a strategic input—like advanced semiconductors—or continue to view it primarily through the climate lens.
Inputs for Decision-Making
When planning the energy mix for AI, two key inputs matter most:
Relative efficiency – total output per unit cost (lifecycle cost of electricity).
Time to production – how quickly capacity can come online:
Any credible plan must balance near-term practicality with long-term energy potential.
The Solutions
History shows the formula: reduce regulations + provide targeted tax credits + fund research. From the telegraph to railroads, radio, nuclear, computing, the internet, and cell phones—the U.S. has mobilized public-private capital behind new technologies before.
But for what objective?
Guarding Against Capitalism’s Blind Spots
The danger is obvious:
Quarterly earnings and election cycles compress decision-making.
Profit-first thinking leads to “cheap now, costly later.”
Idealistic vs pragmatic focus lead to wasted investments in the long-run.
To guard against this, Ai65 strategy horizon has three layers:
Tactical (0–5 years): Natural gas, coal retrofits, utility level solar + advanced storage. Government seeds long-term bets - piloting SMRs (small “modular” nuclear reactors), piloting advanced storage chemistries.
Strategic (6–15 years): Large nuclear, major grid interconnects and continental transmission lines.
Transformational (16–40 years): AI-optimized grids, advanced Nuclear (even Fusion).
Immediate Actions: 2025–2027
What’s realistic within 2 years?
New natural gas plants & LNG expansions (Liquid Natural Gas): Abundant, reliable, cost competitive.
Coal retrofits with carbon capture (so called “Clean Coal”).
Grid modernization pilots: AI-assisted load balancing, microgrids, limited transmission upgrades.
Solar + advanced storage: Sun-rich states.
Small Modular Nuclear Reactors (SMRs): Pilot deployments fast-tracked. At-site Ai SMRs.
Strategic Energy: 2030–2040
By the 2030s, the race for AI leadership will hinge on which nations invested early in the above solutions.
Nuclear (SMRs + Next Generation Nuclear): Long-term, energy-dense, lowest lifecycle cost when scaled.
Continental transmission: Linking sun-rich Southwest, wind-rich Midwest, and hydro/nuclear in the East.
Advanced solar + storage: Moving to dispatchable energy flow.
Transformational Energy: 2040–2065
In the long arc, AI-optimized energy systems will dominate.
AI-optimized grids: Dynamic, self-healing, global load balancing across continents.
Super-grids: International interconnects, geopolitically complex but technologically feasible.
Next Generation Nuclear (Gen IV): Safer, More Efficient, Some Site Based for Dedicated Ai-energy.
Case Study: Nuclear (2026–2040)
Because Nuclear is the highest output and lowest cost per kWh opportunity for the future, the priority here is real. Past objections and issues with Nuclear will be solved or reduced significantly with dedicated effort.
Technology trajectory: Nuclear efficiency gains + higher safety will help make Nuclear the standard for Ai.
Nuclear Generations in Context
Gen I (1950s–1970s): First commercial reactors. Small, early, experimental.
Gen II (1970s–1990s): Most reactors still operating today. Standard light-water reactors (PWRs, BWRs). Reliable but designed with 20th-century needs in mind.
Gen III / III+: Safer, more efficient. Examples: AP1000 (Westinghouse), EPR (Europe), and
the Vogtle 3 & 4 units in Georgia.
Next Generation Nuclear (In development):
A family of advanced designs that aim to be safer, more fuel-efficient, produce less waste, and in some cases generate both electricity and industrial products (eg., hydrogen).
Safety: Many designs use passive safety (cooling systems that don’t require active pumps).
Efficiency: Some designs can use existing nuclear waste as fuel, extending output efficiency and reducing storage problems.
Flexibility: Smaller footprints; some are modular.
Timeline: Most serious next generation Nuclear deployments are expected in the 2035–2045.
Global competition: China, Russia, and Canada are all actively pursuing next generation Nuclear.
Location: Development will be a mix of stand alone or At-site for dedicated Ai-energy.
Key Drivers:
Fuel Utilization (how much energy you get from uranium) - today's reactors use 1% of uranium’s potential; Next gen reactors will use 60% of uranium’s potential. 50x output increase!
Thermal Efficiency (heat → electricity conversion efficiency): Current 33% efficiency; Next Gen 45%
Lifecycle Economics (cost per MWh over decades): Current: $90/MWh; Next Gen: $50/MWh (5 cents/kWh)
2065 (and beyond): Work on next generation Nuclear develops materials science, cooling techniques, logistics systems and safety protocols that enable eventual Fusion Reactors. Fusion: $30/MWh (3 cents/kWh)
(Fusion also offers unlimited fuel source - input is seawater!)
Case Study: Solar + Advanced Storage (2026–2040)
Because solar is already cheap per kWh when available (in the sunshine), its real promise lies in
pairing with advanced storage.
Technology trajectory: Solar efficiency gains: Current: Silicon cells deliver a 21%+ efficiency conversion from sunlight to electricity. New classes of crystalline materials will drive cell conversion to 34%. New material science techniques (stacking) will drive conversion efficiency to 50%.
Utility scale solar is cheap in sunny regions: $30/MWh (3 cents per kWh)Location: Sun-rich states (Texas, Arizona, Nevada, California, Florida)
Best Case: Florida Daytime: $20/MWh (2 cents per kWh); Full day with storage: $3/MWh (3 cents/kWh)
Geopolitical considerations: China dominates solar and storage. (a negative for US trade balance, infrastructure reliance, and technology transfer). Short-term construction jobs argue for some solar with advanced efficient storage in sunny states to help utilities buffer costs as Nuclear evolves. However, in the long-run, solar is not the main focus for Ai. With low Irr, the US will cede the solar market to China. Storage innovation may be an area for U.S. involvement.
Climate and America’s Ambitions
Climate does not disappear—but in the AI era, it becomes subordinate.
Climate costs = externalities: Wildfires, floods, pollution, health costs. When priced correctly, fossil fuels including natural gas, are not as cheap as they appear.
Pragmatism wins: Despite our climate ambitions, which emotionally distract us to protect the planet, in the end, are supplanted by the race for Ai and energy leadership. Pragmatism wins over environmentalism.
Renewables’ role: Strategically, solar plus advanced storage play in sunny states. Renewables are tactical.
The Ai65 Energy 40-Year Horizon
For the U.S. to win the race for Ai, we must win the race for low cost 24/7 baseload energy:
Have the lowest cost and most available energy vs. China. Export our technology to Europe, Middle East.
Build the highest “24/7 reliable” energy flow for US AI infrastructure.
Conclusion
The AI machine is hungry, and energy is its food. For the U.S. to win the Ai race, we must think not in quarters or election cycles but in decades. The energy roadmap must be pragmatic, relentless and clearly recognize the geopolitical importance of energy.
Climate will take a back seat for now. The race for AI leadership demands it. Only over the long horizon, with energy abundance delivered via Nuclear, and Ai enabled innovation delivered via U.S. leadership, may finally human progress, America led, align with our hopeful goals of planetary stewardship.
Ai65 Energy
Ai65 Energy is dedicated to exploring the 40-year arc of energy as a foundational input to artificial intelligence. We help leaders navigate the intersection of strategy, technology, and capital to ensure America’s long-term leadership in Ai.
