There is so much hype surrounding AI, and there is even more capital behind it. AI data centers currently support the US economy, accounting for more than 1% of GDP. So when someone suggests that the race may not have won, or worse still, that the AI race may have ended, it makes people sit in their seats.
That’s what Adam Livingston, author of the Bitcoin era, argues. It’s already carrying over the game: China is moving far further, not overtaking the US, but quietly cornering the one resource that one resource needs most, specifically nuclear power generation.
But how true is this story, and how black and white are things really?
Nuclear Scoreboard: Fact vs. Fiction
Livingston highlights the impressive disparity. China currently builds 16 nuclear power plants, but the US has zeros. He’s not too far in his numbers. As of late 2025, around 30 nuclear reactors are under construction in China, approved repeatedly every year, accounting for almost half of the world’s new builds.
Some analysts say China is aiming to reach 65 gigawatts of nuclear capacity by the end of this year, and is aiming to reach 200 gigawatts (about 10 times growth) by 2040.
In contrast, the US completed the Vogtle 3 and 4 reactors after long delays and cost overruns. Currently, there are no new, large-scale nuclear projects at the groundbreaking stage.
But this is not the big picture. For the first time in years, there is a new plan for the US nuclear. Following recent executive orders and policy reforms, the Westin House has announced its intention to build 10 large nuclear reactors by 2030. Work is expected to begin in the next few years.
However, the hurdles of regulation, public skepticism, and the pure complexity of nuclear construction mean that implementation is not guaranteed, and no actual new construction is yet to be underway.
Energy: A real AI bottleneck?
Livingston raises an important question: Do you underestimate the role of pure energy in AI advancement? Model training and reasoning are greedy for electricity.
Training frontier models like the GPT-4 requires dozens of megawatts, and data-centric electricity demand in the US is projected to more than double in the next decade (78 gigawatts by 2035).
Global data centre energy consumption hit 415 terawatts in 2024 and doubled by 2030, with AI increasing share. So, in theory, countries that can deploy the most stable carbon-free electricity will actually be advantageous in AI races.
China’s approach to industrial policy is direct, top-down and offensive. This allowed us to rapidly increase nuclear structures, but American utilities rely more on upgrades, license expansions, and slow, market-based activities.
However, China is moving forward rapidly, but the US is also focusing on improving efficiency and supplementing its bases with new technologies such as small modular reactors (SMRs) and renewable energy.
Have you finished the AI race yet?
As Livingston argues, “the funeral is already happening”? There doesn’t seem to be much decision on the answer. China’s nuclear expansion is realistic and impressive, and collaboration with AI infrastructure is not overstated. AI relies heavily on continuous, affordable power.
But US leaders and businesses aren’t standing still. New projects, policy movements and investments in both energy and AI are recovering, but so far it is not in line with China’s size and speed.
The benefits of America in basic AI research, chip design, cloud infrastructure and venture funding remain important. Even if data center power is constrained, efficiency innovation, smart grids, and distributed calculations can narrow the gap.
In particular, “energy wars” can be just as important as software and data, but the results depend much more than just the number of nuclear power plants. Livingston’s discussion highlights often overlooked aspects of the global technological struggle, but declares funerals early. The scoreboard is changing, but the AI race is not over yet.
