Will China win the AI race? NVIDIA’s CEO speaks up

Nvidia CEO Jensen Huang told a London AI summit that “China is going to win the AI race,” arguing the country is just “nanoseconds” behind the US and poised to overtake it. He cited energy advantages, research talent, and the risk that US chip export controls will galvanize Beijing to close the gap. Huang later softened his stance, saying US AI could still prevail—but the remarks reignited a high-stakes debate.

The “Yes” case: scale, strategy, and coordination

Greg Slabaugh, Professor of Computer Vision and AI (Queen Mary University of London), points to China’s growing research heft and state-backed industrial policy:

• Research lead in computer vision. At ICCV 2025 in Hawaii—one of the field’s top venues—about half of authors were affiliated with Chinese institutions, compared with ~17% from the US, indicating outsized influence over a core AI domain.

• National strategy + capital. Since 2017, Beijing’s New Generation AI Development Plan has aimed for global leadership by 2030, backed by large, state-guided funds such as the National Venture Capital Guidance Fund (~US$138B).

• Infrastructure and firms. Massive AI compute hubs in Beijing, Shanghai, and Shenzhen support research and commercialization. Huawei, Alibaba, Baidu, and newcomers like DeepSeek are building competitive models and domestic hardware.

• Innovation under constraint. Export controls have pushed Chinese teams to optimize algorithms for local chips—often a catalyst for efficiency gains.

• Data and talent scale. With 1.4B people and huge platforms, China generates unmatched data. It also now produces more STEM PhDs than any other country, feeding a self-reinforcing loop of data → talent → investment.

Bottom line: If trends hold, China’s scale + state coordination could deliver leadership in both development and deployment. Slabaugh stresses that meaningful progress still benefits from open, responsible collaboration, balanced by sensible controls on dual-use tech.

The “No” case: US dominance in chips, compute, and capital

Sean Kenji Starrs, Lecturer in International Development (King’s College London), argues the US lead remains decisive—and widening:

• Market leadership. As of November 2025, all top 10 AI firms by market cap are American, and 37 of the top 50 are US-based. Nvidia recently topped US$5T in value. China counts four firms in the top 50—same as Israel.

• Compute gap. The US controls ~39.7 million petaflops of AI compute—about half the global total. China’s ~400,000 petaflops trails even India’s ~1.2M, largely due to bans on advanced Nvidia/AMD chips.

• Workarounds with caveats. DeepSeek showed innovation under constraints, but reports suggest reliance on stockpiled Nvidia hardware and model distillation from US systems—signs of dependency rather than autonomy.

• Energy and efficiency. Cheaper electricity won’t offset that Chinese AI chips are slower and more power-hungry than Nvidia’s best. Meanwhile, US firms deploy at scale across allied data-center hubs (e.g., UAE, Saudi Arabia) built on conditions that exclude Chinese competitors.

• Trajectory, not snapshot. With unrestricted access to cutting-edge chips and hundreds of billions in capital spending, US firms enjoy a compounding lead. In Starrs’s view, this is a marathon, and China is running with weaker legs.

Bottom line: Expect the US to extend its advantage in compute, chips, and capital, even as China advances in selective domains.

What Huang’s remark really signals

• A stress test of export controls: do they entrench the US lead or accelerate Chinese self-reliance?

• A reminder that AI power rests on compute + chip efficiency + data + talent + capital—pillars that shift slowly.

• A call to recalibrate policy between security-minded guardrails and scientific openness that has historically propelled AI.

US-China AI race: The takeaway

• China: formidable research momentum, national coordination, vast data and talent.

• US: dominant chips and compute, capital markets, and a global infrastructure footprint.

• The outcome isn’t strictly zero-sum: interoperable standards, responsible collaboration, and clear safeguards will shape AI’s real-world impact—regardless of who “wins.”