Stanford AI Index Report Is Out: Just How Strong Is China’s AI?

On April 13, Stanford University’s Human-Centered AI Institute (HAI) released the ninth edition of the annual AI Index Report.

Every year around this time, this report becomes a hot topic in industry discussions. This year’s edition is particularly eye-catching—because the report contains quite a few numbers showing “China ranking at the top.”

But getting excited about rankings and numbers is a common ailment on China’s internet. I suggest everyone stay calm and look at how these numbers were actually calculated.

What Does the Report Say?

Let’s start with the key data points:

High-Impact Patents: China surpassed the US in this metric, ranking first.

AI Model Count: The US still leads in the number of top-tier model outputs, but China’s gap is narrowing.

Enterprise AI Investment: The US still leads significantly, but China’s investment density in certain vertical domains is already not low.

Responsible AI Governance: China ranks lower in this category, mainly because the evaluation framework’s dimensions are biased toward Western contexts.

The report’s conclusion: AI competition is shifting from “absolute US leadership” toward a “US-China duopoly.”

But What’s Behind the Numbers Matters More

Honestly, the patent count metric has limited reference value in my view.

The reason is simple: more patents doesn’t equal technological leadership. Many patents are defensive—filed to protect market share rather than to achieve technological breakthroughs. What’s more, many core technological capabilities can’t be measured by patent counts—like the depth of basic research, the density of top talent, or the maturity of engineering capabilities.

If I had to choose the most reliable metric, I’d pick the number of top-tier model outputs.

Not because this metric is perfect, but because it at least reflects a country’s sustained investment and output capacity in cutting-edge AI research. On this metric, the US still has an advantage, and this advantage is unlikely to be erased in the short term.

But this doesn’t mean China has no opportunities. On the contrary, I think China has its own advantages in application-layer innovation and engineering implementation speed.

Why the Application Layer Might Be China’s Home Turf

I mentioned in an earlier article: AI competition’s first half is about foundation models; the second half is about application landing.

In the first half, the US indeed leads. But in the application layer, China has an advantage that’s hard for the US to replicate: massive user data and abundant application scenarios.

Imagine an AI model being repeatedly tested by hundreds of millions of users across dozens of different scenarios—this training process can’t be simulated in any lab environment.

Of course, this doesn’t mean China can rest on its laurels in the application layer. The importance of basic research can’t be overstated enough. But at least on the AI application track, China isn’t without cards to play.

Returning to Stanford’s report, I think the most valuable part isn’t those ranking numbers, but the report’s judgments on AI development trends—AI is shifting from “laboratory technology” to “infrastructure.” And infrastructure, in the end, comes down to engineering capability, implementation speed, and cost control.

These are precisely things China excels at.

So, after reading this report, my conclusion is: don’t be falsely modest, but don’t be blindly optimistic either. Numbers are just numbers. What really matters is knowing how those numbers were derived and what they actually represent.