What is the state of AI? Is China on its way to winning the race in AI?

What is the state of AI? Is China on its way to winning the race in AI?
  • Homepage
  • >
  • All News
  • >
  • What is the state of AI? Is China on its way to winning the race in AI?

The State of AI, a collaborative effort between MIT Technology Review and the Financial Times, examines the way AI is reshaping the global power landscape. Writers from both publications are going to debate a different aspect of the AI revolution that is transforming global power every Monday over the next six week.

John Thornhill, FT tech columnist and Innovation editor and Caiwei Chen from MIT Technology Review discuss the technological battle between Silicon Valley vs. Beijing.

John Thornhill wrote:

From afar, it appears that China will soon become the AI superpower for the 21st Century.

In the West, we are prone to be enamored with America’s technological leadership in AI, semiconductors, and data centres. Warren Buffett, the legendary investor once said: “Never Bet Against America.” It is true that no country has been able to match America’s “incubator of unleashing human talent” for over two centuries.

China today has the opportunity, the motive and the ability to do the technological equivalent of murder. It may not be wise to make a bet on the ability of society to pool its resources to maximize AI’s effectiveness.

Data highlights trends. China is the leader in AI patents and publications. According to Stanford University’s Artificial Intelligence Index Report, 2025, China will account for 22,6% of all citations by 2023. This compares with 20,9% in Europe and 13% in the US. China accounted for 69.7% all AI Patents in 2023. The US still has a large lead among the 100 top publications (50 in comparison to 34 in 2023), however, its share is steadily decreasing.

The US is also ahead of China when it comes to top AI researchers, though the gap between the two countries is closing. A report by the US Council of Economic Advisers shows that 59% of top AI researchers in the world worked in the US, compared to 11% of them in China in 2019. By 2022, these figures would be 42% and 28.

It is possible that the Trump Administration’s new restrictions on foreign H-1B visa holders will lead to more Chinese AI scientists in the US returning home. China could see its talent ratio shift further to the Chinese side.

In terms of technology, US institutions created 40 notable AI models by 2024, while China produced only 15. Chinese researchers, however, have learned how to get more done with less. Their strongest models for large languages–including DeepSeek-V3 from open-source and Alibaba Qwen 2.50-Max–outperform the most efficient US models when it comes to algorithm efficiency.

China will excel at applying open-source AI models in the future. According to the latest Air Street Capital report, China now has more AI models downloaded each month than does the US. China has already surpassed the US in AI-enabled Fintech, E-Commerce, and Logistics.

Perhaps the most intriguing–and potentially the most productive–applications of AI may yet come in hardware, particularly in drones and industrial robotics. China will continue to be a leader in manufacturing as the field of AI research evolves.

Dan Wang is a tech analyst, author and blogger. BreakneckHe has highlighted China’s strengths in manufacturing processes and the engineering state, even if he also showed the negative effects that this engineering mentality can have on the social realm. He told me that China has grown in many ways, both technologically and economically. But repression exists. It is also getting worse.

Caiwei: I would love to know your thoughts on China’s AI dreams. How far will China’s social engineering hinder its technological aspirations?

Caiwei Chen responds:

Hi, John!

It’s true that the US is still a leader in research frontiers and infrastructure. The term “winning AI” can be used to describe many things. Jeffrey Ding in his book Technology and the Rise of Great PowersThe long-term success of a technology that is general purpose, such as AI, often depends on how far and deep it can be spread in society. China has a great chance to be the winner of this race, even if “murdering” is a little too far! ).

China will continue to be a bottleneck for chips. Export restrictions are limiting access to the top GPUs. This has forced buyers to gray market and labs to repair or recycle banned Nvidia stocks. While domestic chip programs are expanding, there is still a performance gap between the top and the rest of the range.

These same limitations have forced Chinese companies to adopt a new playbook – pooling computing, optimising efficiency and releasing models with open weights. DeepSeek’s V3 training, for instance, only used 2.6 million GPU hours, which is far below US competitors. Alibaba’s Qwen model is now among the top-downloaded open weights in the world, while companies such as Zhipu or MiniMax have developed multimodal video and multimedia models that are competitive.

China’s Industrial Policy allows for rapid implementation of new models. Major enterprises and local governments have already begun implementing reasoning models for administration, finance, and logistics.

Another advantage is education. The major Chinese universities have integrated AI literacy into their curriculum, enabling them to embed skills in advance of the demand on the job market. Ministry of Education also plans to incorporate AI training into all levels of schooling. China is not a “engineering-state” but its relationship to new technology has been shaped by decades of top-down coordination and infrastructure development. This system, which often faces less resistance from the public, allows for large-scale adoption. It is important to note that the use of technology at scale allows for iterative improvement.

Stanford HAI’s 2025 AI Index also found that Chinese respondents were the most positive in the world regarding AI’s future, far more so than the populations of the US and UK. This is striking given the fact that China’s economic growth has been slowing for the first two decades since the pandemic. AI is now seen by many in the government and business as an important spark. The optimism can fuel growth, but the question is whether this can continue through slower economic growth.

The social control is still a part of it, but there are new ambitions. These Chinese AI entrepreneurs are among the most global-minded that I have ever seen. They move fluidly from Silicon Valley hackathons to pitch meetings in Dubai. They are all fluent in English, and they know the intricacies of venture capital. After watching the previous generation struggle with the Chinese label burden, they build businesses that are transnational quietly from the beginning.

China has the potential to shape AI in China and around the world. While speed matters, it’s not the same as superiority.

John Thornhill You can also find out more about the following::

Caiwei you’re correct, speed does not equal supremacy. (And “murder” is perhaps too strong of a term). You’re right, too, to emphasize the difference between China’s open-weight model strength and the US’ preference for proprietary models. It’s not only a battle between different economic models, but also two ways to deploy technology.

Sam Altman admitted this earlier in the year that OpenAI was on the wrong side. This is going to make for an interesting story. Who called it right?

More readings on the US-China Competition

It’s been discussed a great deal about the use of generative AI by people in their everyday lives. This article from The The FTThe visual storytelling team of explores reality

China The FT Reporters ask Nvidia how long it can continue to dominate its Chinese competitors

Toys and companion devices, which are novel applications of artificial intelligence that have gained traction in China but will soon be available to Americans as well, represent a real-world application. The US is also gaining interest in this application of AI. MIT Technology Review Story explored it.

As AI and sanctions shift in China, the once frenetic data center construction has come to a halt. MIT Technology Review Story took a look on the ground at how stakeholders were figuring out what to do.

View Article Source

Share Article
Facebook
LinkedIn
X
UK businesses restart delayed investment after the Budget, and most of those announcements will go unheard
Berkshire heating engineers put the boiler service first as January bill forecasts climb
As Christmas lists get shorter, Ascot home brand says one lasting piece beats five impulse buys
Secret Link