America’s (AI) Power Problem: Why the U.S. Risks Falling Behind China

The U.S. Invented the Future—But Its Power Grid May Not Be Able to Support It

Innovation is American.

Electricity is Chinese.

That’s the new competitive gap.

While China modernized its grid for industrial scale, the U.S., the intellectual engine of the AI revolution, still relies on an aging power system that risks limiting the very breakthroughs it produces. China still cannot match Nvidia’s most advanced chips, but its abundant, inexpensive electricity allows Chinese firms to compensate and scale anyway.

Meanwhile, the U.S. is heading toward needing the equivalent of an entire New York State’s worth of additional electricity every year just to power new data centers.

Right now, the grid cannot supply that. Platocom has been writing about this risk for years:

America may lead the world in frontier AI models and advanced chips, but we are soon running into a far more basic constraint: there may simply not be enough electricity to power the AI boom. Long before today’s headlines, we warned that both traditional and AI-driven data-center growth would expose the limits of the U.S.’s aging power grid, overwhelm already congested data-center hubs, and ultimately force a national rethink of energy policy, permitting, and power grid modernization.

Bottom line: The AI race won’t be won by algorithms alone, but by the countries with the electricity to scale them.

A Warning From Texas: the U.S. Power Crunch Is Already Here

Texas offers the clearest early signal of America’s emerging AI power crisis.Crypto miners alone consumed 14.7 million megawatt-hours in 2024 — about 3% of all electricity on the Texas grid, more than 1 million homes use in a year. Link to our August blog How U.S. Data Centers, Led by Texas, Are Draining Water and Energy Supply.

Now AI is dwarfing even that.

In yesterday's Wall Street Journal, Raffaele Huang and Brian Spegele reported on this issue in the artile “China’s AI Power Play: Cheap Electricity From World’s Biggest Grid.”

China Built a Power System Designed for Scale

As reported in the WSJ, between 2010 and 2024 China added more electricity generation than the rest of the world combined, and last year produced more than twice the electricity of the United States.

Some Chinese data centers pay less than half of U.S. electricity prices.

“In China, electricity is their competitive advantage.”

Mongolia is now called "cloud valley of the grasslands." Photo by Herbert Bieser

Inner Mongolia, once a remote, sparsely populated region, is now filled with thousands of wind turbines, ultra-high-voltage transmission lines, and more than 100 new or planned data centers, forming what officials call a “cloud valley of the grasslands.”

Why China’s Power System Is So Strong (and Cheap)

China built the world’s largest and most cost-efficient power system through deliberate, long-term national planning:

• Since the 1970s, leaders directed state-owned companies to build hundreds of coal plants

• Then came mega-scale hydroelectric, wind, and solar projects

• China built the world’s largest ultra-high-voltage transmission network, enabling power to flow thousands of miles with minimal losses

• It is now constructing 34 nuclear reactors, with nearly 200 more planned

The scale is unmatched:

• 3.75 terawatts of generation capacity — more than double that of the U.S.

• Long-term data-center rates as low as 3¢ per kWh

• Faster permitting and construction cycles measured in years, not decades

China treated electricity as a national strategic asset—and AI is now benefiting from that decision.

China Plans to Widen Its Advantage

Morgan Stanley forecasts:

• $560 billion in grid investment through 2030

• A 45% increase over the previous five-year period

Goldman Sachs projects:

• 400 gigawatts of spare capacity by 2030

• Equal to three times expected global data-center demand

China's "East Data, West Computing" plan is a massive national strategy by the government to address the imbalance between the supply and demand of computing power across the country. 

Under its “East Data, West Computing” plan, China aims to channel western renewables to eastern population centers and interconnect hundreds of data centers into a nationwide compute system, a “national cloud”, by 2028.

Meanwhile, the U.S. Grid Is Aging and Overloaded

While China modernized its grid for industrial scale, the U.S., he intellectual engine of the AI revolution, still relies on an aging power system that could limit the very breakthroughs it produces.

Why the U.S. power grid is struggling:

1. It is old. Most of the grid was built in the 1950s–70s and is nearing or past end-of-life.

2. It is fragmented. The U.S. does not have a unified national grid; it is a patchwork of regional utilities and state regulations, making long-range power transmission extremely difficult.

3. Permitting is slow. Where China can build a major transmission line in two to three years, the same line in the U.S. can take 10–15 years due to multi-layered permitting and local opposition.

4. It is expensive. Data centers in Northern Virginia typically pay 7–9¢ per kWh, more than double Chinese rates.

5. It lacks new generation. Morgan Stanley warns that within three years:

U.S. data centers could face a 44-gigawatt electricity shortfall, equivalent to New York State’s entire summertime supply.

In simple speak: The U.S. is heading toward needing an entire New York State’s worth of extra electricity, every year, just to power new data centers. Right now, the grid cannot supply that.

Microsoft CEO Satya Nadella echoed the concern, saying the company worries it may not have enough power to operate the enormous number of chips it is purchasing.

Cheap Power Helps China Offset Its Chip Weakness

Graphic by Gordon Johnson

China still cannot produce semiconductors that match Nvidia’s top-tier chips, but cheap electricity allows Chinese firms to compensate.

Companies like Huawei, Alibaba, Baidu, and DeepSeek can bundle thousands of domestic chips together to approach the performance of Nvidia systems—an approach that consumes enormous amounts of power.

Chinese firms can afford it. American firms cannot.

The WSJ reports that "this strategy only works because China’s power is cheap and abundant."

AI Growth Is Now an Energy Story

AI models consume massive amounts of electricity:

• Training large models

• Running inference (every chatbot prompt, every image generation)

• Supporting data-center cooling, networking, and storage

By 2030, China’s data centers alone may use as much electricity annually as France.Yet U.S. data centers already consume 45% of global data-center electricity, far exceeding China’s 25%.

The U.S. grid simply cannot keep pace.

Unless Washington modernizes transmission, accelerates permitting, and expands both clean and dispatchable power (power that doesn’t depend on weather), the gap will widen, and U.S. leadership in AI might be constrained not by talent or chips, but by electricity.

Innovation is American.

Electricity is Chinese.

That’s the new competitive gap.

Electricity may define who can run AI, but cloud infrastructure will determine who controls data and digital services. Nowhere is this more consequential than in Africa, where both U.S. hyperscalers and Chinese cloud providers are accelerating investment. And while Africa is a central arena in this contest, it is not the only one, similar dynamics are unfolding across Southeast Asia, Latin America, and the Middle East.

Next week's Platocom blog will unpack how this U.S.–China cloud competition is taking shape across the continent, how it fits into the broader global landscape, and what strategic options African governments have in response.

Thank you for reading.