China has moved much closer to the United States in the global AI race, but the two countries are still competing with very different strengths.
Stanford’s2026 AI Index reportsays the performance gap between the top US and Chinese models had narrowed to just2.7% by March 2026. Yet the US still holds a huge lead in private AI investment, with $285.9 billion invested in 2025 compared with $12.4 billion in China.
That gap matters because AI progress depends on more than benchmark scores. Capital can fund chips, data centres, research teams, cloud capacity and the expensive process of training and deploying frontier systems at scale.
China is no longer a distant second in AI
The model race has tightened dramatically.
Stanford says US and Chinese models have traded places near the top of performance rankings several times since early 2025. DeepSeek-R1 briefly matched the leading US model in February 2025, while the top US model led by only2.7% as of March 2026.
China is also strong on the research side. It leads in AI publication volume, citations and patent grants, while the US retains the lead in higher-impact patents and notable model development.
In 2025, US institutions produced59 notable AI models, compared with China’s 35.
That doesn’t mean China has reached full AI parity with the US. It does mean the old assumption that American labs were comfortably ahead on model capability has become much harder to defend.
The shift is already visible in the open-weight market.Chinese AI models are gaining US users with lower costs and open weights, giving developers more alternatives to proprietary AI platforms.
America’s biggest advantage is still capital
The clearest gap sits in investment.
US private AI investment reached$285.9 billion in 2025 China recorded $12.4 billion over the same period — meaning US private investment was more than 23 times higher
The difference also appears in startup formation. The US had1,953 newly funded AI companiesin 2025, compared with 161 in China.
Money doesn’t automatically produce better models.
But it gives companies more opportunities to train, test, deploy and improve them. It can also support the expensive computing infrastructure needed to serve AI products at enormous scale.
Stanford counts5,427 data centres in the United States, more than ten times the number in any other country, as part of its analysis of the infrastructure supporting AI development.
We think this is where the US advantage remains especially difficult to copy quickly. Algorithmic improvements can narrow benchmark gaps faster than countries can build capital markets, cloud infrastructure and data-centre capacity.
China is making efficiency part of the competition
Chinese labs aren’t simply trying to outspend Silicon Valley.
They are increasingly competing through open-weight models, lower costs and rapidly improving systems.
GLM-5.2 is one example ofhow Chinese open models are challenging US rivals, while Nvidia’s Nemotron 3 Ultra showshow US developers are responding in the open-model market.
That changes the competitive equation.
Businesses don’t always need the smartest model available. A slightly weaker system can be more attractive if it is cheaper, easier to customise and capable enough for everyday coding, customer service, research or automation.
For developers in cost-sensitive markets, that trade-off can matter more than who sits at the top of a benchmark.
The investment comparison needs one important caveat
The $285.9 billion versus $12.4 billion figure comparesprivate AI investment. It is not a complete measure of total national AI spending.
Stanford specifically warns that private investment figures probably understate China’s overall AI investment because the country also uses government guidance funds and other state-backed financing.
The US and China therefore finance AI development differently.
American AI growth relies heavily on private companies, venture capital and hyperscalers. China combines commercial investment with industrial policy and government-backed funding.
So the investment gap is real, but it shouldn’t be read as a complete accounting of every dollar flowing into Chinese AI.
Why this matters for South Africa and Africa
For African companies, the practical question isn’t simply whether the US or China “wins” AI.
It’s which ecosystem can make capable AI affordable, available and useful locally.
Chinese open-weight models could appeal to organisations that want more control over deployment costs or where their data is processed. US companies, meanwhile, benefit from mature cloud platforms, developer tools and enterprise ecosystems already used by businesses around the world.
That leaves African buyers with more choice, but also more questions around cost, security, data governance and long-term platform dependence.
China is narrowing the model gap. The US still has a massive financial and infrastructure advantage.
The next phase of the race will show which advantage matters more.
If Chinese labs can keep producing near-frontier models with far less private capital, how long will America’s investment lead remain decisive?
Is China ahead of the US in artificial intelligence?
Not across every measure.China leads in publications, citations and patent grants, while the US still produces more notable AI models and attracts far more private investment.
How much more does the US invest in AI than China?
US private AI investment reached$285.9 billion in 2025, compared with $12.4 billion in China. That is more than 23 times higher, although Stanford warns that private investment data does not capture all Chinese state-backed AI funding.
Why are Chinese AI models becoming more competitive?
Chinese labs are producing increasingly capable models while competing strongly oncost, efficiency and open-weight availability. That can make them attractive to businesses and developers that value affordability, customisation and deployment control as much as top benchmark performance.
