Nvidia CEO Jensen Huang says the U.S. may lead in AI chips, but China has a significant advantage when it comes to the physical backbone needed for artificial intelligence—especially construction speed and energy capacity.
Speaking with Center for Strategic and International Studies President John Hamre in late November, Huang contrasted how quickly the two countries can stand up large-scale projects.
“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,” Huang said. “They can build a hospital in a weekend.”
That rapid pace is only part of his concern. Huang also pointed to the countries’ energy profiles as AI drives massive new power demand.
China, he noted, has “twice as much energy as we have as a nation, and our economy is larger than theirs. Makes no sense to me.” He added that China’s energy capacity continues to climb “straight up,” while the U.S. remains relatively flat.
On the technology front, Huang emphasized that Nvidia is still “generations ahead” of China in AI chips and the semiconductor processes needed to fuel the current AI wave. But he cautioned against assuming that gap will last forever, saying that “anybody who thinks China can’t manufacture is missing a big idea.”
Despite the challenges he laid out, Huang struck an optimistic note about Nvidia’s trajectory, pointing to President Donald Trump’s efforts to reshore manufacturing and accelerate AI-related investment as a supportive backdrop for the company’s future.
‘Insatiable AI demand’
Earlier this year, Huang drew attention by suggesting China could ultimately win the AI race. He later softened that framing, saying the country was “nanoseconds behind America” in a statement shared on Nvidia’s X account.
Nvidia is one of several major tech players pouring billions into new data centers across the U.S.—a buildout that experts tell Fortune could exceed $100 billion over the next year alone.
Raul Martynek, CEO of DataBank, which partners with large tech companies to construct data centers, said the average facility costs between $10 million and $15 million per megawatt (MW), and even a relatively small data center typically requires around 40 MW.
“In the U.S., we think there will be 5 to 7 gigawatts brought online in the coming year to support this seemingly insatiable AI demand,” Martynek said.
At those cost levels, that implies roughly $50 billion on the low end and as much as $105 billion on the high end devoted to new AI-ready data center capacity in the near term.
