Nvidia CEO Jensen Huang says China holds a structural infrastructure advantage that the U.S. can no longer ignore, warning that the next phase of the AI race will hinge not only on chips but also on construction speed and energy supply.
His comments add a sharper edge to a debate already intensifying inside Washington and across the global tech sector: as demand for AI supercomputing explodes, the world’s biggest economies are now racing to see who can build fast enough — and power those systems at scale.
In a late November conversation with Center for Strategic and International Studies President John Hamre, Huang laid out a blunt comparison.
Register for Tekedia Mini-MBA edition 19 (Feb 9 – May 2, 2026): big discounts for early bird.
Tekedia AI in Business Masterclass opens registrations.
Join Tekedia Capital Syndicate and co-invest in great global startups.
Register for Tekedia AI Lab: From Technical Design to Deployment (next edition begins Jan 24 2026).
“If you want to build a data center here in the United States from breaking ground to standing up a AI supercomputer is probably about three years,” he said. China, he added, can erect major infrastructure at a stunning pace. “They can build a hospital in a weekend.”
Those remarks tap into a long-running story about China’s industrial model. For two decades, the country has run what is arguably the world’s fastest large-scale building system: highways constructed at double-digit mileage per day, megacities carved out of farmland in under five years, and full manufacturing zones built in the time it takes U.S. developers to secure local permits. Much of this comes from China’s highly centralized planning structure, its tolerance for round-the-clock construction, and a regulatory environment that moves state-approved projects forward with few of the environmental-review delays common in the U.S.
That speed, Huang warned, is about to matter more than ever. The AI boom has forced companies and governments to think in gigawatts rather than gigabits. Massive clusters of GPUs require vast campuses, power lines, cooling systems, transformers, and highly specialized electrical infrastructure. If the U.S. wants to compete in the next generation of supercomputing, it must build all of that much faster than it currently can.
Energy Capacity: Huang Says China Holds Another Advantage
Huang pointed to another imbalance: national energy supply. China, he said, has “twice as much energy as we have as a nation, and our economy is larger than theirs. Makes no sense to me.” According to him, while U.S. energy capacity has stayed “relatively flat,” China’s continues to grow “straight up,” reinforcing its lead in the long-term ability to power AI supercomputers.
Energy is emerging as the new bottleneck in the AI race. U.S. utility companies have begun warning that they cannot keep up with the skyrocketing electricity demand from data centers. Some regions are already issuing moratoriums on new power-hungry facilities. Some analysts say permitting delays, aging transmission lines, and the slow pace of new power-plant construction all threaten to derail ambitious AI expansion plans in the U.S.
China, meanwhile, has poured trillions of yuan into grid expansion projects over the years, from sprawling ultra-high-voltage transmission lines to rapid power-generation buildouts led by state-owned giants. That includes coal, natural gas, nuclear, hydro, and solar deployments — all growing simultaneously, often at speeds not seen elsewhere.
Nvidia Still “Generations Ahead,” but Huang Warns Against U.S. Complacency
Despite the infrastructure gap, Huang stressed that Nvidia remains “generations ahead” of China in AI chips — the essential engines for training and operating large AI models. He said Nvidia’s semiconductor lead remains intact, particularly in the advanced manufacturing techniques required for high-performance GPUs.
Still, he delivered another caution: the U.S. should not underestimate China’s ability to catch up. “Anybody who thinks China can’t manufacture is missing a big idea,” he said, a reference to China’s rapid capacity to scale hardware production once political and economic priorities align.
Huang’s remarks also came after he briefly stirred controversy in early November when he predicted China would win the AI race. He later softened that framing on his company’s X account, saying China was “nanoseconds behind America,” signaling a race too close — and too fast-moving — for any clear long-term winner.
He remains optimistic about Nvidia’s direction, citing President Donald Trump’s push to reshore manufacturing and accelerate AI investments. Those policies, he suggested, could give the U.S. a counterweight to China’s speed and energy advantages.
A High-Stakes Building Boom
Nvidia is one of the most aggressive players in a U.S. data-center buildout that industry executives say will surpass anything seen before. Experts tell Fortune the country could see over $100 billion in new construction in the next year alone.
Raul Martynek, CEO of DataBank, which builds data centers for major tech firms, said one facility costs roughly $10 million to $15 million per megawatt (MW). A smaller-sized center typically requires 40 MW. Scaling those numbers up, he said, the U.S. is preparing to bring 5 to 7 gigawatts of new capacity online next year — a staggering expansion.
At those prices, the U.S. is preparing for a spending wave between $50 billion and $105 billion. The growth is fueled by cloud giants, model labs, and AI startups racing to secure capacity as training cycles grow more computationally intensive.
The scale of the boom also shows why Huang keeps drawing attention to China’s structural advantages. If Beijing can deliver the same scale of capacity in half or a third of the time — and with significantly more energy available to power it — the geopolitical balance in AI infrastructure could tilt quickly.