The AI Energy Race: Why America’s Power Grid Crisis Could Cost It the Lead Over China

The United States has focused its competitive attention on semiconductors, export controls, and algorithm development in its contest with China over artificial intelligence supremacy. But experts across finance, diplomacy, and energy policy are raising an alarm about a dimension of that competition that has received far less attention: the electricity needed to actually run the AI being built. On that front, China is pulling ahead — and the gap is widening faster than most people realise.

Contents

1. The energy dimension of the AI race — why it matters
2. America’s power grid: a system under acute strain
3. China’s energy advantage: three decades in the making
4. US vs China: the energy scorecard
5. The clean energy market: a $7 trillion economic play
6. Can the US close the gap — and what would it take?
7. Where the two countries might find common ground

The energy dimension of the AI race — why it matters

Training and operating large AI models consumes extraordinary amounts of electricity. A single AI-related computation can draw up to 1,000 times the energy of a standard web search — and the number of those computations being performed globally is growing exponentially. The data centres that house the servers performing those calculations require reliable, abundant, and increasingly affordable power to remain economically viable. Without it, AI development faces a hard physical constraint that no amount of algorithmic ingenuity can circumvent.

The United States entered this era with a genuine energy advantage: it is effectively energy independent, meaning it does not rely on foreign suppliers for its domestic fuel needs. But energy independence and energy abundance are not the same thing. American utilities are increasingly unable to keep pace with surging demand, electricity prices for consumers are climbing, and communities across the country are pushing back against the infrastructure needed to accommodate massive new data centre facilities. The result is a growing gap between the electricity that AI development requires and the electricity that the grid can reliably supply.

The core risk: PJM Interconnection, the largest electricity grid operator in the United States, serving over 65 million people across 13 states, projects it will be six gigawatts short of its own reliability requirements by 2027. Analysts at Gartner forecast that power shortages will constrain operations at 40% of AI data centres by that same year.

America’s power grid: a system under acute strain

For roughly two decades before the AI surge began, United States electricity demand was nearly flat — growing at well under 1% annually. The grid was built for that reality. The sudden arrival of hyperscale data centre development has overturned those assumptions almost overnight. Today the entire US data centre sector draws under 15 gigawatts of power from the grid. The pipeline of facilities currently under construction or approved would, if fully built, add close to 150 gigawatts of peak demand — an increase of nearly 20% on the country’s current total electricity usage.

The infrastructure response to that demand is already running into hard physical limits. Grid transformer lead times have stretched to three to five years, creating backlogs that are delaying new data centre connections across the country. Some utility providers have paused all new data centre interconnections entirely while they work through their existing queues. One utility, AEP Ohio, has halted all new connections due to insufficient power infrastructure. In states including Georgia, Indiana, Missouri, and Washington, local communities are demanding that technology companies fund their own power plants and transmission upgrades rather than passing those costs onto ordinary ratepayers.

The electricity that has been added to the US grid in recent years has come predominantly from solar and wind installations — the only two energy sources that can be deployed at meaningful scale within the timeframes currently required. But current federal policy is pulling back from both, creating a paradox where the demand for new clean electricity is accelerating while the support framework for building it is contracting.

A Department of Energy analysis warns that if current power plant closure plans proceed alongside AI-driven demand growth, blackout frequency across the United States could increase by a factor of 100 before the end of the decade. The states served by the PJM grid — including Maryland, Pennsylvania, and Virginia, which host a disproportionate share of the country’s data centre infrastructure — face the sharpest projected exposure.

China’s energy advantage: three decades in the making

China’s position in global clean energy did not emerge from a sudden strategic pivot toward artificial intelligence. The country’s systematic investment in renewable energy technology began in the late 1980s and early 1990s, when it began inviting Western firms to transfer wind turbine and solar panel technology in exchange for access to the Chinese domestic market. One five-year plan after another kept renewables central to national energy strategy — through administrations, through economic cycles, and well before AI became a geopolitical priority.

The result of that accumulated investment is now visible at a scale that has no precedent in energy history. Since 2021, in just five years, China has added more total power generation capacity across all energy technologies than the United States has constructed across its entire history. In the first half of 2025 alone, China installed 357 gigawatts of new wind and solar capacity — more than the entire installed power capacity of India. Its solar installations in that period were more than double the combined additions of every other country on earth.

China has also built the transmission infrastructure to move that electricity where it is needed. Ultra-high-voltage transmission lines now connect renewable generation sites in the country’s remote western interior to data centre clusters and industrial demand in the heavily populated east — a project of staggering engineering scale that has no equivalent elsewhere in the world. The country’s “East Data, West Computing” initiative has embedded this logic into national planning, concentrating new data centre construction in regions where land and renewable electricity are abundant and affordable. In May 2026, China activated its first large-scale project to supply renewable energy directly to a data centre, a 500-megawatt solar installation in Ningxia linked directly to a cloud computing facility in the city of Zhongwei.

The East Data, West Computing strategy: China’s government is deliberately co-locating data centre development with its most abundant renewable energy sources — remote inland regions with high solar and wind resource intensity. The result is a national AI infrastructure that is geographically distributed, grid-connected, and increasingly powered by cheap, clean electricity, with AI computing costs in China now estimated at below $3 per million tokens, compared with roughly $15 per million tokens in the United States.

The clean energy market: a $7 trillion economic play

China’s clean energy strategy has never been motivated purely by environmental ambition. The country’s planners have consistently treated renewable energy technology as an industrial and export policy — a sector in which early, sustained investment could generate dominant global market share and the economic returns that flow from it. That calculation is proving correct.

EV exports from China grew by 80% in 2025. Battery exports climbed 40%. Solar panel exports rose 20%. The country now controls approximately 80% of global solar panel and battery technology manufacturing, and over 70% of wind turbine production. The global clean energy market is projected to reach $7 trillion annually by 2035, and China has positioned itself to command a substantial share of that demand. Clean energy is, as observers have noted, both green for the environment and green for economic returns — and the Chinese government understood that conjunction decades before it became widely acknowledged.

The export numbers speak clearly: China’s clean energy exports were worth approximately $76 billion in 2024. The country’s manufacturing capacity across solar, wind, and battery technology is producing a compounding economic return that makes continued domestic investment in renewable deployment simultaneously an environmental, strategic, and commercial decision — each reinforcing the others.

Indonesia Data Center: Why Billions Are Flowing Into Southeast Asia’s Digital Hub

Can the US close the gap — and what would it take?

The honest answer to whether the United States can compete with China across the full spectrum of clean energy technology — solar panels, wind turbines, batteries, and large-scale grid infrastructure — is almost certainly no, at least not in the near term. The cumulative investment gap is measured in trillions of dollars, the manufacturing capacity gap reflects decades of strategic prioritisation, and the transmission infrastructure gap would require years and substantial federal commitment to meaningfully close.

What the United States can do is make the choices that prevent the gap from becoming permanently decisive. New technologies crossing from innovation to manufacturing to commercial deployment face a cost and risk hurdle that private markets alone are generally unwilling to clear without policy support. This is precisely the lesson of the rare earth mineral supply chain, where the Trump administration has acknowledged that competing with China requires adopting elements of the Chinese industrial playbook — targeted government investment to de-risk early-stage manufacturing capacity and create secure domestic markets for new technologies.

html

The Inflation Reduction Act, passed under the Biden administration, represented the most serious attempt in American history to close this gap — a mix of subsidies, tax incentives, and domestic manufacturing requirements that attracted significant inward investment from foreign firms seeking access to the US market. Most of those provisions have since been unwound. Whether private market economics can substitute for that policy framework in driving clean energy deployment at the required scale and speed is one of the defining open questions of the current political moment.

Why the World Needs Data Centers: The Invisible Engine of Our Modern Lives

Where the two countries might find common ground

For all the competitive tension in the US-China technology relationship, there is at least one area where serious analysts on both sides see the potential for cooperation rather than confrontation: the safety and governance of AI systems themselves. The risks posed by advanced AI that is inadequately understood, poorly controlled, or deliberately weaponised are not national risks — they are civilisational ones. An accident or a catastrophic misuse of AI would not respect geopolitical borders.

The trust deficit between Washington and Beijing is real and substantial — in some respects wider today than at any point in recent memory. But the argument made by senior figures with deep experience of both countries is that the alternative to finding a framework for coexistence and selective cooperation is a world that becomes more dangerous and less prosperous for both. Competition and collaboration are not mutually exclusive — and the domain of AI safety, where both countries have an interest in preventing catastrophic outcomes, represents one of the more credible starting points for rebuilding the diplomatic architecture that would make that dual relationship sustainable.

The bottom line: The United States leads China in AI model capability, data centre scale, and chip design. China leads the United States in clean energy capacity, electricity affordability for AI workloads, and long-term energy infrastructure investment. The country that resolves its constraint first — either by closing its energy gap or by pulling ahead decisively on AI capability — will have a significant structural advantage in the decade that follows. Right now, the energy constraint is the United States’ problem to solve.