AI infrastructure demand is set to accelerate sharply into 2026–2027, and Korea is being positioned as a critical supply-chain and robotics partner in that expansion.

NVIDIA Corporation(NVDA.US) CEO Jensen Huang, during his latest visit to Seoul after a seven-month gap, outlined a clear message: AI infrastructure scaling is speeding up, memory demand remains tight, and Korea will play a growing role not just in chips—but in robotics and physical AI.

AI infrastructure: “Second half will be much larger”

Huang described the current phase as an accelerated build-out of AI systems, driven by strong momentum from existing architectures and next-generation platforms.

He highlighted that:

• The Grace Blackwell systems have already delivered strong performance in the first half of the year
• The Vera Rubin architecture is now in full production
• Supply-chain coordination is becoming increasingly important as scale expands

He also emphasized the expected step-change in demand:

“The scale in the second half of this year will be substantially larger than in the first half, and next year’s scale will be very, very large.”

He added that NVIDIA’s goal in Korea is to ensure partners are fully aligned as demand for components such as DRAM and HBM continues to rise.

Memory supply: NVIDIA denies cutting HBM usage

One of the most closely watched topics during the visit was high-bandwidth memory (HBM) supply.

Huang directly rejected speculation that NVIDIA had reduced its use of high-speed memory in its systems.

“We will use a large amount of high-speed memory.”

He acknowledged that supply constraints remain, but stressed that allocation will be managed more efficiently across systems rather than reduced.

He also confirmed that all three memory suppliers supporting the Vera Rubin platform have passed qualification and are now in mass production. Samsung Electronics and SK Hynix (Related Asset: iShares MSCI South Korea ETF(EWY.US)) together account for roughly 70% of NVIDIA’s AI chip memory supply. 

Korea’s role: from semiconductors to “physical AI”

Beyond semiconductors, Huang placed strong emphasis on robotics as a long-term growth area for Korea.

He argued that Korea’s combination of:

• Advanced manufacturing
• Semiconductor expertise
• AI capabilities

creates a strong foundation for robotics and “physical AI” applications.

“Because Korea is a global manufacturing hub, we can directly apply robotics and physical AI technologies… to local industries.”

He added that future semiconductor manufacturing itself will increasingly rely on robotics and AI-driven systems, opening further collaboration opportunities with local chipmakers.

Analysts in the region also noted that Korea’s industrial base makes it well positioned for this shift toward embedded AI systems across factories, vehicles, and robotics platforms.

Expansion in Korea: R&D hiring and future campus plans

Huang confirmed that NVIDIA is already expanding its presence in Korea beyond supply-chain partnerships.

Key developments include:

• Active recruitment for a local R&D center
• Plans to eventually build a full campus in Korea
• Deeper collaboration with Korean AI and robotics talent

He noted that Korea’s talent base and manufacturing ecosystem make it an ideal location for long-term research investment.

Business coordination and “surprise” hints

During meetings with major Korean conglomerates including Hyundai Motor, LG, SK, Samsung Electronics, and Naver, Huang said the focus was on keeping the supply chain aligned for large-scale AI production.

He also hinted at additional announcements:

“I’m bringing substantial business opportunities for Korea. Moreover, I’ll also be unveiling some surprises.”

However, he did not provide further details.

On the ground: meetings, esports, and public appearances

Beyond formal business discussions, Huang’s itinerary included informal engagements that drew public attention.

He visited a Seoul internet café, met esports figures including Lee Sang-hyeok (“Faker”), and was accompanied by his daughter throughout the trip. He is also scheduled for a ceremonial baseball first pitch and a TV talk show appearance.

Bottom line

NVIDIA is signaling a clear acceleration in AI infrastructure deployment into 2026 and beyond, with:

• Stronger demand expectations for AI compute systems
• Continued reliance on high-bandwidth memory supply chains
• And a strategic expansion of Korea’s role into robotics and physical AI development