US and China: Nvidia CEO on the AI Infrastructure Race

The CEO of Nvidia, Jensen Huang, has warned that the US risks falling behind China in crucial areas for AI infrastructure, namely the speed of construction and national energy capacity.

During a discussion with John Hamre, President of the Centre for Strategic and International Studies, Jensen highlighted the differing paces at which each nation can establish large-scale digital infrastructure.

He also questioned whether the US possesses the necessary power resources to satisfy the growing demands of high-level computing. His remarks arrive at a time when data centre operators are forecasting major expansion across the US, with billions in investment anticipated to fuel the rapid growth of AI workloads.

Construction speed and energy concerns

Jensen explained to John that the disparity in construction timelines is emerging as a critical element in the development of AI infrastructure.

“If you want to build a data centre here in the United States, from breaking ground to standing up an AI supercomputer is probably about three years,” says Jensen. he adds: “[China] can build a hospital in a weekend.”

He also voiced concerns regarding the comparable national energy availability, suggesting that China's capacity is considerably larger than that of the US, at a time when data centre operators are rushing to secure long-term power agreements.

China has “twice as much energy as we have as a nation and our economy is larger than theirs. Makes no sense to me”, he says. Jensen observed that China’s energy capacity is continuing to increase sharply while growth in the US has remained relatively static.

For operators that are planning multi-gigawatt campuses, having predictable access to power is becoming as vital as obtaining semiconductors.

The US chip advantage

Despite the infrastructure challenges he outlined, Jensen stressed that the US holds a distinct advantage in AI chip technology. He stated that Nvidia remains “generations ahead” of China in the design and manufacturing processes that form the foundation of modern AI systems.

However, he advised against complacency.

Jensen adds that “anybody who thinks China can’t manufacture is missing a big idea”.

His comments could reflect a wider apprehension within the semiconductor and data centre ecosystem: even with a technological lead, the US must also have the capability to deploy infrastructure at the required scale and speed to support its rapidly expanding AI economy.

Jensen also alluded to the political momentum in the US surrounding reshoring and AI investment. He said President Donald Trump’s economic priorities could bolster domestic production and future infrastructure development, presenting a route to fortify national capabilities.

Investment surges to meet AI demand

Nvidia is one of several major technology firms making substantial investments in new data centres across the US as AI workloads increase. Industry analysts predict that the scale of new construction activity will grow substantially in the coming year.

Raul Martynek, CEO of DataBank, said that the rising demand for AI compute is reshaping investment strategies throughout the sector.

“In the US, we think there will be 5 to 7 gigawatts brought online in the coming year to support this seemingly insatiable AI demand,” he says, according to Fortune.

Raul estimates the cost of a data centre to be between US$10m and US$15m per megawatt. With a smaller facility typically needing around 40MW, capital expenditure is on the rise, especially for operators planning expansions across multiple sites.

The forecast of 5 to 7GW of new capacity translates to approximately US$50bn to US$105bn in new construction activity. These numbers highlight the magnitude of the challenge that data centre developers are facing.

They must balance the swift pace of AI adoption with the construction capacity, energy availability and supply chain resilience required to deliver some of the most extensive digital infrastructure projects ever undertaken in the US.