The statements by Jensen Huang, CEO of Nvidia, at the FT Future of AI Summit (where he went so far as to state that “China will win the AI race” before clarifying that it is just “nanoseconds behind”) have once again put an uncomfortable question on the table: are they changing the rules of the game in favor of Beijing?
The race for AI supremacy is more than a business competition; It is a matter of state. The technology giant has issued a clear warning about technological advances in the Asian giant and how these could redefine global leadership in the coming years.
The new battle: energy and infrastructure
Huang summarized the central idea: the AI competition today is largely an infrastructure race. Three factors weigh in on its diagnosis: energy costs, regulatory agility and physical scale of data centers.
China can offer cheaper, subsidized electricity in key regions, and is promoting large data center projects that concentrate capacity on a national scale. These operational differences reduce the marginal cost of training and launching very large models.
Added to this is regulatory fragmentation in the West: multiple states and jurisdictions make homogeneous deployments difficult and lengthen deadlines, while the centralized Chinese approach allows for faster decisions and investments (something that for Huang translates into strategic advantage).
Signs of Chinese advance
Sanctions and chip export restrictions have had a double effect. On the one hand, they limit immediate access to cutting-edge hardware; On the other hand, they push the Chinese industry to optimize and develop local alternatives, accelerating efficiency solutions and its own chips.
That dynamic explains why observers see China advancing not just in quantity but in pragmatic implementation of AI.
Furthermore, although China starts from a lower aggregate capacity base than the US, the country consolidates large regional projects to shorten latencies and scale production inference. That is, it is committed to applying AI on a large scale, not just publishing research models.
The counterpoint: the technical and creative advantage of the West
Even so, the West retains relevant levers. Recent studies estimate that the United States accounts for about 74% of global high-performance computing capacity for AI, a material advantage when it comes to training the largest and most expensive models.
That computing supremacy, along with ecosystems of capital, talent and academic freedom, remain strengths that are difficult to replicate in the short term.
The question is to what extent the traditional computing advantage offsets Chinese operational efficiencies: cheaper energy, state coordination and massive deployments.
An open, multi-speed race
Huang’s words are not a definitive verdict, but rather a strategic warning: competitiveness in AI is no longer decided only in laboratories and papers, but in who builds, regulates and powers the infrastructure at scale.
China combines speed of implementation and public policies that reduce costs; The US maintains leadership in computing and foundational innovation.
The result will depend on political decisions (energy, trade and regulation), investments in infrastructure and the ability of each block to convert capacity into useful products.
In that race, as Huang suggested, every nanosecond and every kilowatt counts.
This post is also available in: