Energy: The Overlooked Force Powering the AI Explosion

Why cheap, abundant power—not just chips or models—will decide winners in the next decade

AI's hunger for electricity is reshaping global energy demand at a scale few appreciate. Massive data centers, GPU clusters, and inference at billions of queries per day require unprecedented power generation and storage. Batteries paired with renewables are emerging as the fastest path to unlocking grid capacity, while long-term visions point to space-based solar compute. Political and regulatory hurdles slow progress in the West, but the physics and economics favor rapid scaling—especially for integrated players.

Key Takeaways

• AI data centers consume city-level electricity; cooling alone can eat 40% of usage, with inference scaling to billions of daily queries driving even higher demand.
• Batteries + solar effectively double grid capacity by storing off-peak energy and discharging during peaks, bypassing the need for many new power plants or transmission lines.
• Tesla deployed 46.7 GWh of energy storage in 2025 (up ~50% year-over-year), with energy gross margins around 29-31%—far outpacing automotive margins.
• Battery costs have plummeted from over $1,000/kWh in 2010 to ~$115 today, following Wright's Law patterns of 20-30% cost drops per production doubling.
• Distributed home batteries (like Powerwalls) form virtual power plants, stabilizing grids and enabling resilience without centralized failures.
• Nuclear offers reliable baseload for AI, with co-location deals accelerating, but permitting delays (15-20 years) favor faster-deploy solar + storage.
• China leads in solar and battery rollout; the U.S. faces grid bottlenecks, NIMBY opposition, and regulatory delays.
• Space-based solar AI compute—enabled by low-cost launches, free cooling in vacuum, and constant sunlight—could become the lowest-cost path for massive inference in the coming years.
• The full ecosystem convergence: solar/batteries for power, rockets for deployment, AI for workloads, and satellite networks for connectivity.

The AI Energy Crunch Is Already Here

Training large models and running inference at global scale demands enormous electricity. A single large training run can match a small city's consumption, while widespread adoption—every search, image generation, agent task, or autonomous vehicle query—multiplies that need exponentially. Cooling adds massive overhead, often 40% of total draw. The result: energy and AI are no longer separate challenges. The companies securing reliable, low-cost power gain decisive edges.

Batteries Unlock Grid Capacity Without New Plants

The grid wastes vast potential because it lacks storage—power plants run for peaks but idle much of the day. Batteries charge during low-demand periods (cheap renewables) and discharge during spikes, effectively adding capacity without new generation or major transmission builds.

This approach scales dramatically at utility levels. Tesla's Megapack units store enough to power thousands of homes briefly, but fleets create transformative effects. In 2025, deployments hit 46.7 GWh globally, with quarterly records like 14.2 GWh in Q4. Revenue from energy generation and storage reached ~$12.8 billion, up significantly year-over-year.

Why Margins Tell the Story

Energy storage delivers higher profitability than cars. Recent quarters show energy gross margins in the high 20s to low 30s, driven by scale and manufacturing efficiencies. This profitability gap highlights where the real leverage lies as demand surges.

Wright's Law Drives the Cost Flywheel

Cumulative production doubles bring predictable cost declines—historically 20-30% for lithium-ion batteries. From $1,000+/kWh a decade ago to around $115 now, further drops to $80 or below appear likely soon. Cheaper batteries make solar viable around the clock, crushing intermittency complaints and enabling massive renewable builds.

Home Batteries Create Distributed Resilience

Powerwalls store 13.5 kWh—enough for half a day or more of typical home use when paired with solar. Millions connected form virtual power plants: during grid stress, coordinated discharge stabilizes supply, pays owners, and reduces blackout risks. Over a million units already save homeowners billions annually. This distributed model builds redundancy and flexibility the old centralized grid cannot match.

Nuclear's Role: Reliable but Slow

Nuclear delivers dense, carbon-free baseload—ideal for 24/7 AI needs. France relies on it for ~70% of power. Co-locating plants with data centers cuts transmission losses. Tech giants pursue these deals. Yet U.S. builds take 15-20 years due to permitting, costs, and overruns. Small modular reactors promise speed, but deployment lags. In contrast, solar + batteries scale incrementally and faster.

The Long Game: Solar from Space

Earth receives more solar energy in one hour than humanity uses yearly. In orbit, panels capture 6-8x more due to no atmosphere, night, or clouds. Run AI compute directly in space—radiate heat to vacuum for "free" cooling—and beam results down via satellites. Launch costs, already slashed by reusability, drop further with next-gen systems, making space data centers potentially cheaper than terrestrial ones facing grid, permitting, and cooling hurdles.

This vision ties solar expertise, battery tech, low-cost orbital access, and AI workloads into one system. Recent ecosystem moves underscore the convergence.

Political Roadblocks vs. Global Race

China installs solar and batteries at unmatched speed—more in single years than much of the world combined. Deserts offer ideal solar sites; wind corridors abound. Yet U.S. progress stalls on oil/utility lobbies, NIMBY fights against farms/turbines/lines, slow regulators, and environmental lawsuits. Grid upgrades lag: mid-20th-century design struggles with bidirectional flows from millions of sources. Transmission expansions face 10-15 year timelines.

The bottleneck is not technology or economics—it's will and policy. AI demand forces action, but delays risk ceding leadership.

Tesla's Integrated Advantage

One company spans generation (solar), storage (home/grid-scale), vehicles (mobile batteries), and AI optimization. Energy already outpaces automotive in margins and growth velocity. As AI drives demand, this vertical integration captures value across the stack—powering models, stabilizing grids, and enabling scale.

Abundance or Constraint Ahead

Energy abundance is inevitable—the physics works, costs fall, demand explodes. The path matters: deliberate acceleration yields leadership; delays create chaos and lost ground. The next years decide who powers the AI era.