Nvidia CEO Jensen Huang recently highlighted a significant advantage that China holds over the United States in the realm of artificial intelligence (AI) infrastructure. Speaking at a forum with John Hamre, president of the Center for Strategic and International Studies, Huang emphasized China’s strengths in construction and energy, particularly in how swiftly it can complete large-scale infrastructure projects.
“Building a data center in the United States, from the initial groundbreaking to the deployment of an AI supercomputer, typically takes about three years,” Huang remarked. In contrast, he pointed out that China can erect a hospital in a single weekend, noting the stark differences in infrastructure development timelines.
Huang’s concerns extend beyond just construction speed; he also expressed apprehension regarding the comparative energy capabilities of the two nations. He stated that China possesses “twice as much energy as we have as a nation,” and despite the U.S. having a larger economy, its energy capacity remains relatively stagnant while China’s continues to expand rapidly.
Despite these concerns, Huang reiterated Nvidia’s leadership in AI chip technology, asserting that the company is “generations ahead” of its Chinese counterparts in this field. Nevertheless, he cautioned against complacency, suggesting that underestimating China’s manufacturing capabilities could be a grave mistake.
Huang’s insights come at a time when the demand for AI technology is skyrocketing. Just recently, he modified previous claims about the AI race, initially suggesting that China might emerge as the victor, only to later clarify that the nation is merely “nanoseconds behind America” in the competition.
Nvidia is not alone in its investments; numerous tech giants are committing substantial resources to data center construction in the U.S., with projections indicating that these investments could surpass $100 billion in the upcoming year. Raul Martynek, CEO of DataBank, noted that the average cost to construct a data center ranges from $10 million to $15 million per megawatt (MW), with smaller facilities generally requiring around 40 MW. Martynek forecasts that the U.S. could see 5 to 7 gigawatts of new energy capacity coming online in the following year to meet the ever-growing demand for AI technologies. This projection translates to an estimated cost of $50 billion on the lower end and potentially up to $105 billion on the higher end.
