South Korea and NVIDIA Open a Joint AI Lab and Deepen Tech Ties at San Francisco Summit
A new research lab at one of Asia's top universities, a memory chip deal, and plans for AI infrastructure signal South Korea's push to become a serious player in global AI.
Key points
- NVIDIA and KAIST (Korea Advanced Institute of Science and Technology) announced a joint AI research lab in Seoul on 24 July 2025, the first between a Korean university and any global tech company.
- South Korean President Jae Myung Lee attended the AI Summit in San Francisco alongside the country's top business leaders and researchers.
- SK Group and NVIDIA expanded an existing partnership in June 2025 to co-develop memory chips for AI hardware.
- SK Telecom announced plans in June 2025 to build AI infrastructure supporting physical AI and robotics in Korea.
South Korea is putting its flag firmly in the AI race. At a summit in San Francisco this week, President Jae Myung Lee sat down with chip giant NVIDIA and a group of Korean business leaders to map out what the country's AI future could look like. It builds on NVIDIA founder and CEO Jensen Huang's visit to Korea last month.
The headline announcement: NVIDIA and KAIST are opening a joint AI research lab at the KAIST Kim Jaechul Graduate School of AI in Seoul. NVIDIA reported the news directly.
Why does this lab matter?
It's the first joint AI lab between a Korean university and any global technology company, which signals that South Korea wants to build AI expertise at home rather than simply buy the finished product from abroad.
The lab focuses on agentic AI, meaning AI systems that can carry out multi-step tasks on their own rather than answering a single question. Picture an AI that books your flights and emails the hotel without you clicking anything. Researchers will use NVIDIA's Nemotron open models, a family of AI models NVIDIA releases publicly for researchers to build on, plus cloud computing from NVIDIA's partners.
What about the business side?
Korea's corporate giants are moving too. SK Group Chairman Chey Tae-won joined Huang for dinner in Woodside, California the evening before the summit, alongside teams from SK Hynix and SK Telecom. It's the kind of relationship-building that tends to precede serious money.
SK Hynix makes the high-bandwidth memory chips that power AI processors. In June the two companies announced an expanded deal to co-develop memory for NVIDIA's AI hardware platforms, covering data-centre AI and personal AI devices alike. SK Telecom said separately in June that it plans to build infrastructure for physical AI, which covers robots and smart machines that interact with the real world.
| Partner | Announcement | Date |
|---|---|---|
| KAIST | Joint AI research lab in Seoul | July 2025 |
| SK Hynix | Co-develop memory chips for NVIDIA platforms | June 2025 |
| SK Telecom | Build AI infrastructure for robotics and physical AI | June 2025 |
For ordinary Koreans the practical upshot is jobs and research funding built closer to home. For everyone watching from outside, the more interesting bet is that owning a slice of the AI supply chain, from chips to software to academic research, is smarter than depending on one foreign partner. NVIDIA has been busy spreading those bets globally: we reported on 24 July that NVIDIA took a 9.3% stake in European AI cloud company Nebius, and AMD is pressing hard on the chip side too.
The summit continues, with more meetings between Huang and President Lee expected.
Common questions
What is KAIST?
KAIST stands for Korea Advanced Institute of Science and Technology. It's a government-funded university in Seoul widely regarded as one of the best science and engineering schools in Asia.
What does 'agentic AI' mean in plain English?
Agentic AI is software that plans and completes a series of tasks by itself, say researching a topic and drafting a document, without a human steering each step.
Does this affect consumers outside South Korea?
Not directly, not yet. Research labs like this one often produce the models and tools that eventually show up in apps people use every day, so the work done here could matter down the line.
