The AI Race Is About Building Power in Time

Why V2G Is America’s Real-Time Path To Winning. Can We Win Without It?

Data centers, driven overwhelmingly by AI, will require 134 gigawatts of power by 2035, according to S&P Global. That’s roughly 25% of current U.S. generating capacity, built in five years. Everyone agrees the demand is real. The question is whether we can actually meet it.

This isn’t an article about the cheapest solution or the most elegant engineering. It’s about the only solution that can actually be deployed in time, and why getting there first might be the most consequential economic decision of the decade.

Why Winning This Race Actually Matters

As we approach artificial general intelligence, model performance isn’t improving linearly, it’s increasing exponentially. Each generation of AI is dramatically more capable than the last. But here’s the inflection point that should focus minds; once models learn to teach themselves recursively, improvement becomes self-accelerating. Whoever is in second place at that moment may find the gap impossible to close.

We’re leaving the Mad Max world, the era where hydrocarbons are the effective currency of geopolitics, where controlling oil meant controlling power. We’re entering a transition period where abundant, cheap electrons gate the AI race. Beyond that lies a world where compute itself becomes the dominant currency. The nations that can train the most powerful models, run the most inference, and iterate fastest will set the terms of the next economic era.

Many of us still think about battery supply chains in terms of global markets for key materials like lithium. But in a head-to-head competition with an adversary that controls over 90% of global lithium processing, that assumption becomes a structural vulnerability. China has already demonstrated its willingness to throttle supply of strategic materials, as it recently did with rare earth elements. Acquiring sufficient control of lithium and other strategic metals essential to winning this race, without the specter of Chinese supply disruption, is a key locus of both risk and opportunity going forward.

This is where the EV in your driveway becomes unexpectedly relevant. V2G doesn’t just solve a grid problem, it democratizes participation in the compute economy. If America wins this race, those batteries become a distributed asset that lets ordinary citizens participate in the infrastructure of AI. Keep your car plugged in and your AI is effectively free, you are trading electrons for compute. If we lose, we become customers with limited influence rather than owners of the next technological revolution, masters of our own destiny.

The Power Problem Has a Deadline

According to the EIA’s latest data, the U.S. has only 18.7 GW of new combined-cycle gas turbines planned through 2028, with just 4.3 GW currently under construction. Meanwhile, GE Vernova, one of only three major turbine manufacturers globally, has a 50 GW backlog with deliveries not expected until late 2028. Wood Mackenzie reports the industry is already running at 90% of global manufacturing capacity.

The math is simple: we need over 130 GW, and the entire global gas turbine supply chain can deliver maybe 20 GW in the timeframe that matters. Nuclear takes 15 years. Fusion is real and getting closer, but commercial deployment won’t arrive until the late 2030s. New gas can’t be built fast enough.

The 2,000 GWh Asset Hiding in Driveways

By 2030, the United States will have approximately 30 million electric vehicles on the road. At an average battery capacity of 60 kWh, that’s 1,800 GWh of mobile storage, more than all the pumped hydro storage ever built in America.

This capacity already exists. It doesn’t need to be manufactured, financed, or permitted. It just needs to be connected.

Vehicle-to-Grid technology turns EVs into distributed batteries. During the day, they charge from solar at workplaces. In the evening, they discharge to power homes and feed the grid during peak demand. The cars are parked 90% of the time anyway; V2G puts that idle capacity to work.

How the Capacity Model Works

The key metric is Effective Load Carrying Capability (ELCC), how much of your generating capacity can be counted as firm power that’s reliably available. Solar panels alone get only 5% ELCC because the sun doesn’t shine at night. Add 4 hours of battery storage and ELCC rises to 35%. Add another 4 hours from V2G, for 8 hours total, and you are at 55%.

What about the remaining 45–65%? That’s where existing gas peakers come in. The U.S. already has 500 GW of gas capacity with about 150GW of gas peakers running at 5–10% capacity, there to provide guaranteed power reliability. If we have a V2G network at play, we don’t need to build an unachievable number of new turbines; we just need the existing fleet to provide backup during extended cloudy periods and multi-day weather events.

At roughly 25% fleet participation, 30 million EVs provide 450 GWh of grid-accessible storage, enough to add 4 hours of duration to existing grid batteries for over 100 GW of data center demand. Combined with traditional storage, this gets us to the 55% ELCC range, with existing gas handling the rest.

The key point: even with only one in four EV owners participating, and without data centers paying car owners for battery access, which would be a private transaction, not a government cost, the model pencils out. That leaves considerable upside as participation rates grow and compensation mechanisms mature. Ultimately a world where you trade compute for battery access makes this a zero-sum transaction for both sides of the trade.

The Practical Question: What About My Morning Commute?

This is the first question everyone asks, and it has a straightforward answer. The average American commute is 40 miles round trip. Modern EVs have 200–300 miles of range. V2G systems allow owners to set minimum charge thresholds, to never discharge below 60%, for example.

A 60 kWh battery at 60% charge still holds 36 kWh, enough for roughly 130 miles, more than three times the average commute. The grid gets access to the top 40% of battery capacity during evening peak hours (4–8 PM), then overnight charging partially refills the battery when electricity rates are lowest.

The operational model is simple: charge from solar during the workday, discharge during evening peak at home, recharge overnight from the grid. EVs are parked 90% of the time. V2G makes that idle time productive.

The Myth That’s Costing Us the Race

The conventional wisdom is that V2G will destroy EV batteries. The data tells a different story.

Battery degradation is dominated by calendar aging, the slow chemical breakdown that happens whether you use the battery or not. This accounts for 85–90% of total degradation over a battery’s life. Active cycling adds only 10–15% at baseline. V2G cycling increases this to 20–25%, adding roughly 9–14% additional degradation over 10 years.

CATL now offers 15-year, 1.5 million kilometer warranties on EV batteries with a 0.2% claims rate. BYD provides 8-year warranties with 3,000+ cycle ratings. These aren’t speculative projections, they are contractual guarantees backed by billions of dollars in liability. Bottom line- we are now in a world where the car wears out before the battery does.

The Economics: Battery Cost Eliminated

The breakthrough in V2G economics is simple: the batteries are already paid for.

Traditional grid storage requires $650/kW in battery capex. That’s money that has to be recovered through electricity sales. But EV batteries were purchased for transportation, their capital cost is already covered and the trade of compute for power works for both buyer and seller. No need for a government subsidy at all.

With battery capex eliminated, V2G + solar delivers power at $42–46/MWh. Compare that to new CCGT plants: $43–49/MWh at today’s $4/MMBtu gas prices, rising to $64/MWh when gas hits $7/MMBtu , which most analysts expect by 2030–2035.

At current gas prices, V2G is roughly break-even with new gas. At projected gas prices, V2G saves $20+/MWh. But here’s the point that matters: you can’t build the gas plants in time anyway. The supply chain is maxed out. V2G is competitive AND it can actually be deployed in time.

China’s Head Start

While the U.S. debates whether V2G is viable, China is ready to deploy it at scale.

China ended 2024 with 31.4 million EVs on the road , 9% of its vehicle fleet. By 2030, projections show 130–150 million EVs, representing roughly 30% of China’s 420 million vehicles. The government has launched 30 V2G pilot projects across major cities including Beijing, Shanghai, Shenzhen, and Guangzhou, with Shenzhen alone testing thousands of connected vehicles.

The U.S. will reach 30 million EVs by 2030, about 11% of its fleet. China will have 4–5 times as many vehicles at nearly 3 times the penetration rate. More critically, China has the policy framework, the grid software, and the utility coordination that the U.S. still lacks. Currently, only 1% of U.S. chargers are bidirectional.

The asymmetry runs deeper than fleet size. America operates on a 4 to 8 year planning cycle where each subsequent administration often reverses the progress of the prior, frequently on cultural rather than economic grounds. China operates on a 100-year planning horizon that is relentlessly focused on economics. This is how they lifted 800 million people out of poverty in under two decades. When it comes to energy infrastructure that requires decade-long commitments, one of these approaches has a painfully obvious structural advantage.

China has fleet, software, and policy working together. The U.S. has fleet only.

The 36-Month Sprint

V2G deployment requires three things: bidirectional chargers, grid integration software, and utility interconnection agreements.

The technical standards exist. The regulatory pathway is clear; FERC opened wholesale markets to distributed resources back in 2020. Bidirectional chargers are in production. BMW, Hyundai, Nissan, and Volkswagen already offer V2G-capable vehicles. The hardware is ready.

What’s missing is deployment velocity. The U.S. needs to install bidirectional chargers at scale, sign utility interconnection agreements, and build out the aggregation software that coordinates millions of vehicles. None of this requires new technology. It requires decisions.

The Bottom Line

When you need 134 GW of power soon and the gas turbine supply chain can deliver 20 GW, the options narrow quickly. Nuclear is too slow. Fusion isn’t ready. New gas can’t be manufactured. When the critical bottleneck in your supply chain is lithium and 90% of your supply comes from your AI race competitor China, you have a very big problem. A problem that demands that you not let 30 million EV batteries simply sit on the sidelines.

V2G isn’t perfect. It requires new infrastructure, new utility relationships, and new ways of thinking about the grid. But it has one overwhelming advantage: it can actually be built in the time we have.

The batteries are already in driveways. The standards are already written. The economics work, especially as gas prices rise. The question isn’t whether V2G is the optimal solution. The question is whether we’ll deploy the only solution that can be deployed in time to win the AI race.

There is little question that whoever wins the AI race sets the global direction. Compute replaces hydrocarbons as the dominant currency, what Peter Diamandis frames as silicon replacing carbon. What we don’t know, and it is the existential question of our time, is whether we end up in the dystopian world of the Dark Enlightenment popular with some in the venture community, or whether the utopian world of Abundance Theory wins out. Being the dominant leader in AI at least gives the opportunity to choose.

Either way, the coefficient of bureaucratic drag is going to make the transition very painful for global democracies, a subject receiving far less attention than it deserves. The nations that can’t generate cheap electrons at scale will find themselves permanently on the wrong side of an expanding capability gap.

Democratization of compute, having electrons in your driveway that can participate in the infrastructure of AI, is one way to tip the scales toward a positive outcome. V2G isn’t just an energy solution. It’s a vote for a future we can control.

— — —

Figures in this article are drawn from publicly available EIA, S&P Global, Deloitte, Wood Mackenzie, and OEM disclosures. Full assumptions, calculations, and source documents are available in the working paper at TheLastProspector.com.

If you think this analysis is wrong, incomplete, or relies on flawed assumptions, I invite you to challenge it directly. Substantive counter-arguments and alternative models are welcome and will be published alongside the working paper on the website.

Steve Parry is a geologist-turned-venture capitalist who writes about energy transition and resource economics at TheLastProspector.com

Sources

U.S. EIA Preliminary Monthly Electric Generator Inventory (December 2024)

S&P Global Data Center Power Demand Forecast (2024)

Deloitte AI and Data Center Power Forecast (2024)

Wood Mackenzie Gas Turbine Manufacturing Capacity Analysis

China Ministry of Public Security NEV Registration Statistics

Applied Energy V2G Battery Impact Study

Supporting economic model: ELCC, storage duration, and gas sensitivity analysis included in the working paper