US–CHINA AI RACE · CHIPS AND POWER

America has the chips. China has the power.

AI needs two things in huge amounts: chips to compute with, and electricity to run them. The US leads on the first and has cut China off from the best chips; China makes far more electricity and adds more every year. Elon Musk says power will decide it. This site tracks whether that balance is shifting, with data every morning.

In three lines

  1. Power: China generated 2.3x as much electricity as the US in 2025, and in that one year added 491 TWh, more than all the world's data centres use (415 TWh).
  2. Chips: of all the AI compute shipped worldwide in 2025, only 4.7% went to China. The largest US training run used 50x the compute of China's largest.
  3. Yet China's best model is only about 4 months behind. China makes up with domestic chips (87% of its new compute) and with power: its AI data centres use about 2.4x the electricity for the same compute.

The premise we check

Holding

The US leads on chips, China leads on power

  • Chips: China got 4.7% of the AI compute shipped worldwide in 2025, and the largest US training run used 50x the compute of China's. US lead holding.
  • Power: over the latest 12 months China's generation grew by 568 TWh, the US's by 128 TWh (4.4x). China's lead holding.

How this is decided is on the Method page.

The two bottlenecks

Compute is limited by how many good chips you have, and by how much power you can plug them into. Each country is strong in one.

POWER China ahead

2.3x

China's generation vs the US · 2025

  • Added in a year: China 491 TWh, US 129 TWh
  • New capacity: China 532 GW, US 44 GW
  • But not cheaper: coastal data centres pay about what Virginia pays
Power page →

CHIPS US ahead

4.7%

China-bound share of new AI compute · 2025

  • Per chip: B300 = 3.3x Ascend 910C
  • Largest training run: US 50x China
  • Domestic chips now 87% of China's new compute
Chips page →

Where it shows: the models

Despite 50x less compute for its biggest runs, China's best model trails by months, not years. Most of China's best models are open for anyone to download (6 of its top 10); none of the US top 10 are.

Best US modelGPT-6 AstraECI 166.6 · Sep 2026
Best Chinese modelKimi K3ECI 157.7 · Jul 2026
China is behind by4.4 monthsUS models first reached that level this many months earlier

Models page →

Is anything changing?

1 of 6 signals have moved. The one that has already changed is China's switch to domestic chips: US export controls now reach only a small part of what China buys.

All signals and the timeline →

Latest events

  1. ChipsChina

    China's industry ministry reportedly tells ByteDance, Alibaba and others it intends to approve purchases of Nvidia's RTX Pro 5500 (US export approval still unclear) Implicator, China signals it will approve Nvidia RTX Pro 5500 purchases (28 Sep 2026, following The Information)

  2. ChipsBoth

    Trump and Xi meet in Washington: an AI dialogue is set up, but no deal on chip controls Roll Call, AI export controls debate rages as Trump, Xi meet (23 Sep 2026)

  3. ChipsChina

    China reportedly to allow fewer than 200,000 H200s (under half of requests), for training only TrendForce

  4. ChipsUS

    Commerce: the licence rules also cover overseas arms of China-headquartered companies Al Jazeera, US says ban on AI chip shipments applies to Chinese firms outside China (1 Jun 2026)

  5. ChipsUS

    Commerce licenses H200 sales to about 10 Chinese firms (up to 75,000 each) TrendForce

Full timeline →

What the insiders say

Jan 2026 · Power and chips

Elon MuskTesla, SpaceX, xAI

China will win on power, sort out the chips, and far exceed the world in AI compute

China's going to have more power than anyone else and probably will have more chips. Based on current trends, China will far exceed the rest of the world in AI compute. People are underestimating the difficulty of bringing electricity online.

What the data sayOn power he is right in direction, though not yet at 3x: in 2025 China generated 2.34x as much as the US. On compute the data point the other way so far: China's share of new AI compute fell from 11% in 2023 to 4.7% in 2025.

Watch: China's generation vs the US, and whether China's compute is catching up — Chips

Source: Business Insider, Elon Musk says China will 'far exceed the rest of the world in AI compute' (7 Jan 2026, on Moonshots with Peter Diamandis)

Nov 2025 · Power and chips

Jensen HuangNvidia

China will win (later: China is 'nanoseconds behind'); energy in China is almost free

China is going to win the AI race. … (statement hours later) As I have long said, China is nanoseconds behind America in AI. It's vital that America wins by racing ahead and winning developers worldwide.

What the data say'Nanoseconds' is closer to months: China's best model (Kimi K3) sits at a level US models first reached 4.4 months earlier. On price, OIES finds Chinese data-centre power costs US$ 0.08–0.14/kWh on the coast, near Virginia's US$ 0.10; the cheap power comes from local subsidies of up to 50%.

Watch: How many months apart the best models are, and whether Chinese power is really cheap — Models

Source: CNBC, Nvidia's Jensen Huang softens his 'China will win the AI race' remark to FT (6 Nov 2025)

Jan 2026 · Chips

Dario AmodeiAnthropic

The US is many years ahead in making chips; selling them to China is a big mistake

We are many years ahead of China in terms of our ability to make chips, so I think it would be a big mistake to ship these chips. It's a bit like selling nuclear weapons to North Korea.

What the data sayPer chip, the best US design (B300) does 3.3x the work of China's best (Ascend 910C). In 2025 all China-bound chips together added 636,000 H100s' worth of compute, against 12.9 million elsewhere.

Watch: The per-chip gap and the gap in compute shipped — Chips

Source: Axios, Anthropic CEO on Nvidia chips to China (20 Jan 2026, Davos)

Jun 2025 · Power and chips

Ren ZhengfeiHuawei founder

Chips are a generation behind, but clustering many of them closes the gap in practice

Our single chip is still a generation behind the US. We use mathematics to make up for physics, non-Moore's-law methods to supplement Moore's law, and cluster computing to make up for single chips.

What the data sayThe data bear out the method and its cost: China's big AI data centres use about 2.3 MW per 1,000 H100s' worth of compute, against 0.9 MW in the US, roughly 2.4x the power. Clustering works because China has the power to spare.

Watch: How much more power the same compute takes — Chips

Source: ChinaTalk, Huawei founder on US v China and basic research (translation of the People's Daily interview, 10 Jun 2025)

All 11 voices →

Related sites

This page organises information; it does not recommend buying or selling anything. Data belongs to each publisher and may contain errors or delays. Quotes are taken from reports and transcripts; see the sources for the originals. Make your own investment decisions.

Updated 29 Sep 2026 12:24 JST · US power to Jun 2026, China power to May 2026 · models to 9 Sep 2026