Inside Nvidia and Hyundai's Big Bet on Humanoid Robots

Saemangeum: A $7 Billion AI Valley Takes Shape

The most tangible outcome of the Huang-Chung meeting is the Saemangeum project. The reclaimed land area on South Korea's west coast is being positioned as the country's future "AI Valley"—a concentrated industrial zone combining compute infrastructure, robotics manufacturing, and clean energy .

During the meeting, Chung formally asked Huang to co-invest in the project. Huang accepted, announcing Nvidia's intention to build a dedicated AI data center on the site . Huang also discussed opening an Nvidia AI robotics research center in North Jeolla Province to anchor the development .

The numbers are substantial. Hyundai has committed 9 trillion won—approximately $6.3 billion to $7 billion—to the Saemangeum cluster, with 5.8 trillion won allocated specifically for the AI data center and the remainder for South Korea's first dedicated robot manufacturing plant . When complete, the facility is expected to serve as Hyundai's primary production base for Atlas humanoids and other Boston Dynamics robots, with the data center providing the compute needed to train and run the simulation environments that underpin Hyundai's physical AI strategy .

The Atlas Roadmap: From CES Prototype to Assembly Line

Hyundai's relationship with Boston Dynamics—which it acquired for $1.1 billion in 2021—is now moving into an operational phase . While the Seoul meeting didn't originate the Atlas mass-production timeline, it served as the moment when the roadmap locked into place with a major AI infrastructure partner.

Here is the deployment sequence as confirmed by the companies: Boston Dynamics began producing the new all-electric Atlas at its Boston headquarters in early 2026, with the first fleets already committed to Hyundai's Robotics Metaplant Application Center and to Google DeepMind this year . By 2028, Atlas will begin performing parts-sequencing tasks at Hyundai Motor Group Metaplant America in Georgia . The long-term capacity target, announced at CES 2026, is 30,000 Atlas units per year .

Hyundai plans to buy "tens of thousands" of robots—Atlas humanoids, Spot quadrupeds, and Stretch logistics robots—and integrate them across its global manufacturing network . The stated rationale is a shift toward "human-centered automation," where robots handle repetitive, high-risk tasks while humans move into oversight and higher-skill roles .

The Broader Physical AI Partnership

Beyond the data center and the Atlas timeline, Huang and Chung broadened their existing alliance to cover what they called the full physical AI stack. The partnership now spans autonomous mobility, AI-powered manufacturing, and robotics .

Nvidia had already signed a contract to supply Hyundai with 50,000 Blackwell GPUs, and the companies agreed during the Seoul talks to deepen that deal and expand co-development on software-defined vehicles, digital twin factory simulations, and robot training infrastructure .

Huang told reporters the two companies are "very very close" to industrializing robotics, adding that they plan to bring AI to "all forms of mobility" . The language marks a shift in ambition: Hyundai isn't just a customer for Nvidia's chips. It's becoming the manufacturing-scale proving ground for Nvidia's entire physical AI and Omniverse simulation stack.

Huang's Korea Tour: Securing the Memory Supply Chain

The Hyundai summit was the headline event, but Huang's four-day South Korea trip—which began June 5 after appearances at Computex and GTC Taipei—was primarily about locking in the memory chip supply chain that powers Nvidia's AI platforms . He moved through a carefully choreographed schedule of meetings with the country's most powerful conglomerates.

SK Hynix and SK Group. The most consequential semiconductor announcement came from Huang's dinner with SK Group Chairman Choi Tae-won and SK Hynix CEO Kwak Ro-jung at Kkanbu Chicken in Gangnam on June 7 . Huang confirmed that Nvidia's new Vera CPU will use SK Hynix DRAM, and the companies announced a multi-year technology partnership to advance HBM4 memory . Both sides said they are preparing for large-scale collaboration in the second half of 2026 and into 2027 . Huang also warned that the current memory chip shortage could persist for several years .

Samsung Electronics. Huang met Samsung leadership to discuss the HBM supply chain, including potential deeper collaboration on capacity and next-generation memory technologies . Samsung, alongside SK Hynix and Micron, has been qualified to supply HBM4 chips for Nvidia's Vera Rubin AI platform .

SK Telecom. The company plans to use Nvidia's DGX platform to build a gigawatt-scale AI cloud in South Korea, with the first AI factory expected online in 2027 . A formal partnership agreement was signed during the trip.

LG Group and Naver. Huang also met LG Chairman Koo Kwang-mo and Naver Chairman Lee Hae-jin to discuss AI cooperation in consumer electronics, AI services, and cloud infrastructure .

Korea as a Physical AI Beachhead

Throughout the trip, Huang framed South Korea in strategic terms. He called the country "one of the top AI countries" and identified robotics as its next major growth sector . The logic is straightforward: South Korea's memory chip dominance makes it essential to Nvidia's hardware supply chain, while its manufacturing infrastructure and Hyundai's global factory footprint make it the natural place to industrialize physical AI.

The June 2026 meetings effectively completed a pivot that began during Huang's previous October 2025 visit, when he first met Samsung's Lee Jae-yong and Hyundai's Chung over fried chicken and beer and pledged 260,000 GPUs to Korean industry and government . What was then a supply commitment has now become a joint building project—a physical AI supply chain stretching from HBM4 memory fabs in Icheon and Pyeongtaek to a reclaimed-land AI Valley in Saemangeum, ending in Atlas robots handling parts on assembly lines in Georgia.