Is China Catching Up with the US in AI?

Catch-up is real in software and open-model ecosystems, but not uniform. Export controls, advanced fabrication, and capital markets still favor the US for absolute frontier training. Expect a contested middle: competitive Chinese models, constrained chip supply, and intensifying policy rivalry.

The US–China AI race shapes markets, standards, alliances, and which stacks global enterprises can safely adopt.

• Competitive Chinese model releases

  Domestic labs ship models that surprise on price-performance for many tasks.

  Signal · strong

• Chip access constraints

  Leading-edge accelerator supply remains the binding constraint for some training runs.

  Signal · strong

• Application-layer adoption in China

  Local products integrate AI quickly across consumer and industrial use cases.

  Signal · moderate

• Policy and standards contest

  Both sides treat AI as strategic infrastructure, not just a commercial sector.

  Signal · strong

• Scenario A

  Narrowing software gap

  Chinese models approach parity on many benchmarks despite hardware limits.

• Scenario B

  Bifurcated stacks

  Two semi-separate AI ecosystems form with limited cross-adoption.

• Scenario C

  US lead re-widens

  Next-gen compute and labs compound advantage faster than catch-up efforts.

• OpenAI
• Google
• DeepSeek
• Huawei
• Nvidia
• SMIC

• research

  Comparative model evals across US and Chinese releases

• policy

  Export control and semiconductor policy updates

• news

  Domestic chip and model funding announcements

1. Sep 22, 202660%
2. Sep 15, 202658%
3. Sep 8, 202656%
4. Sep 1, 202654%

• Will Open-Weight AI Models Catch Up?
• Can Nvidia Maintain AI Chip Dominance?
• How Close Are We to AGI?