Nvidia placed itself at the center of South Korea's all-out AI push as Huang headlined the government-hosted AI summit in San Francisco before joining President Lee Jae Myung and Korea's business leaders for a relaxed waterfront dinner over beer and fish and chips, replacing Seoul's trademark chimaek but preserving the easy camaraderie that has come to define Huang's relationship with Korea.
At first glance, Korea's chip ecosystem was the centerpiece. Samsung Electronics and SK hynix dominate advanced memory, while the country combines world-class telecommunications, electronics manufacturing, automobiles, robotics and industrial production into one of the world's most complete AI supply chains.
But beneath the celebratory mood lay an anxiety confronting the world's AI champion.
Nvidia's growth rests on a simple equation: more data centers mean more graphics processing units sold, more high-bandwidth memory installed and more networking equipment deployed.
That equation is increasingly running into a wall in the United States.
The constraint is no longer chips or capital. It is electricity.
Power has become the scarce resource in the AI race. Utility interconnection queues in major U.S. markets stretch for years, community opposition has slowed new projects, and transformers and other heavy electrical equipment remain in short supply. Even developers with financing and Nvidia GPUs in hand often cannot connect new facilities to the grid.
Research by Sightline Climate estimates that 30 percent to 50 percent of large data-center projects expected globally this year could face delays because of power shortages, equipment bottlenecks and permitting hurdles.
For Nvidia, every delayed data center represents GPUs that cannot be shipped, networking systems that cannot be installed and memory chips that remain unsold. A bottleneck in Virginia or Texas eventually shows up in demand forecasts in Santa Clara.
Korea as Nvidia's second growth engine
That is where South Korea enters—not as a beneficiary of Nvidia's expansion, but as an alternative engine of demand that America's grid cannot yet accommodate.
The plan calls for roughly 18.4 gigawatts of AI data-center capacity by 2035. SK Group is expected to develop around 15 gigawatts, GS Group 2.4 gigawatts and Naver about 1 gigawatt, representing one of the largest coordinated AI infrastructure programs outside the United States.
Layered on top is Seoul's sovereign AI initiative, encouraging Korean companies to build large language models optimized for the Korean language, domestic regulations and industrial applications.
For Nvidia, those projects provide exactly what U.S. infrastructure constraints increasingly threaten: sustained demand.
"South Korea has all the ingredients to become a global AI powerhouse—world-class networks and data centers, leadership in chip technology and vast industrial scale," Huang said.
"Together with SK Telecom and SK hynix, we are building a new generation of AI factories that will power Korea's next wave of growth."
Partnerships built for mutual gain
The agreements announced during Lee's Silicon Valley visit illustrate why Nvidia's relationship with Korea extends well beyond supplier and customer.
SK Telecom plans to build AI factories totaling up to two gigawatts using Nvidia's next-generation Vera Rubin computing platform while sourcing HBM4 memory from SK hynix. The first facility is expected to come online in 2027.
The economics are compelling.
Servers—primarily GPUs and memory—account for roughly 60 percent of the capital cost of a large AI data center. Every hyperscale AI factory therefore lifts Nvidia's processor sales while simultaneously expanding demand for SK hynix's advanced memory.
"In the AI era, competitiveness depends not just on how effectively AI is utilized, but on how much intelligence we can produce," SK Group Chairman Chey Tae-won said. "We will help Korea evolve from an AI adopter into a global AI innovation hub."
Hyundai Motor Group is pursuing another dimension of the partnership.
Rather than simply buying chips, Hyundai plans to integrate Nvidia's Blackwell GPUs and Drive platform across autonomous vehicles, robotics and smart factories, turning Korea into a proving ground for physical AI.
"Nvidia's planned strategic investment and our infrastructure supply agreement with Brookfield have propelled Naver's vision for the AI factory business into a robust execution phase," Naver founder and Chairman Lee Hae-jin said.
"Leveraging our partnerships with global technology leaders, we will strengthen Korea's sovereign AI ecosystem and enhance the country's AI competitiveness."
Together, the projects reveal Nvidia's broader strategy.
Rather than relying overwhelmingly on U.S. hyperscalers, the company is cultivating national AI champions capable of building infrastructure at scale. Korea has become the most advanced example of that model.
A golden era—if the lights stay on
The week's grandest announcements came from conference stages, but its defining image came around a dinner table overlooking San Francisco Bay.
Lee greeted Huang by joking that he had seen reports of the Nvidia chief enjoying chimaek during his Seoul visit and wished he had joined him.
Huang laughed and replied that everyone in Korea seemed to love AI—from fried chicken restaurants to barbecue houses.
He noted that Nvidia and Korea had worked together for more than 25 years, first through PC gaming and now through artificial intelligence.
"From AI chips and physical AI to AI infrastructure, Korea has achieved remarkable results over the past year," Huang told Lee. "This is truly the beginning of Korea's golden era."
Lee's response was brief.
He hoped it was only the beginning.
Korea needs Nvidia's processors, software ecosystem and AI leadership to accelerate its ambitions.
Nvidia, meanwhile, increasingly needs Korea's memory leadership, industrial manufacturing, coordinated investment and, above all, the electrical capacity to deploy the computing power that America's own grid is struggling to absorb.
The relationship is no longer one-sided.
Korea may have every layer of Huang's AI cake.
Nvidia is helping bake it—but it also needs somewhere to serve it.
Copyright ⓒ Aju Press All rights reserved.