AI · Oct 28, 2026
RTX Spark laptops get open speech models and a dial for background agentsNVIDIA’s Seoul showcase is about owning Korea’s robot and agent developer base
· Updated · Signal Desk newsroom
Analysis of reporting by NVIDIA Blog. How we work

AI Day Seoul drew more than 2,000 people and a who’s who of Korean robotics and telecom firms. Read alongside NVIDIA’s new government partnership, it looks less like a conference and more like ecosystem building.
Key takeaways
- More than 2,000 attendees gathered over two days at COEX for sessions on physical AI and AI agents.
- NVIDIA’s Deepu Talla said Korea is “punching above its weight,” citing Doosan Robotics, Hyundai Motor Group, LG Electronics and SK Group.
- More than 500 developers took Deep Learning Institute workshops and certification exams in Korean.
- Hackathon finalists built Korean-language agents on the OpenShell runtime; winner FlyGate received a DGX Spark.
- Ten Inception startups pitched, including TwelveLabs, Neubility and Avikus.
A recap with a strategy behind it
NVIDIA’s write-up of AI Day Seoul is an event recap, and it reads like one: attendance figures, sponsor names, a happy quote from an SK Telecom engineer. But the shape of the event matters more than the adjectives. Over two days at COEX, more than 2,000 people attended sessions on two themes NVIDIA is pushing hardest right now, physical AI and AI agents. Deepu Talla, NVIDIA’s vice president of robotics and edge AI, named Doosan Robotics, Hyundai Motor Group, LG Electronics and SK Group, plus startups Real World and Luda Robotics, as Korean companies using NVIDIA platforms for general-purpose robotics.
Context makes the purpose clearer. Days earlier, NVIDIA and Korea’s science ministry widened their sovereign AI partnership to cover chips, data centers, talent, startups, models, agents and security. AI Day Seoul is the developer-facing half of that deal: the place where the talent and startup commitments become workshops, certifications and pitch sessions.
Training the next generation on NVIDIA tools
The numbers that tell us the most are about developers. More than 500 took full-day Deep Learning Institute workshops and certification exams in Korean, covering inference deployment at scale, robotics workflows on the Isaac platform and secure AI agents, including a new AI Model and Application Development exam. A community day drew groups such as PyTorchKR, Python Korea, Hugging Face KREW and Womxn Who Code Korea. Hackathon finalists, co-hosted with Fastcampus, built agents trained on Korean language and culture using NVIDIA’s OpenShell runtime.
That is smart and entirely rational ecosystem building. Developers who learn on Isaac, OpenShell and DGX Spark tend to keep using them, and certifications create a steady supply of engineers fluent in NVIDIA’s stack. It mirrors the approach NVIDIA took in Germany’s industrial AI push. For Korea, the benefit is real skills and access to tools. The cost is deeper dependence on one vendor’s platform at a time when the ministry partnership also invited Korean NPU makers into the conversation.
Who showed up, and who did not say much
Breakout speakers came from Coupang Intelligent Cloud, LG AI Research, NAVER Cloud, SK Telecom and Upstage, a list that covers most of Korea’s serious AI builders. Ten Inception startups pitched at a physical AI session, from video-understanding company TwelveLabs to autonomous-navigation firms Neubility and Avikus. An AI for Science forum brought in the Korea Institute of Science and Technology Information and NVIDIA Research.
What the post does not offer is any concrete deployment number, contract or product announcement. Talla’s line about Korea punching above its weight is a compliment, not a metric. We would look to the follow-through, robots shipping in Hyundai or LG facilities, Korean-language agents in production at SK Telecom or NAVER, for evidence that the event translated into something measurable.
Who it affects
- Korean developers
- Free or low-cost training and certification in Korean on NVIDIA’s robotics and agent tools.
- Korean robotics and telecom firms
- Closer ties to NVIDIA’s physical AI platforms, with the lock-in that implies.
- Korean chip startups
- Domestic NPU makers compete for the same developer attention NVIDIA is cultivating.
What to watch
- Concrete deployments by Doosan, Hyundai, LG or SK tied to NVIDIA robotics platforms.
- Progress on the chips, data center and talent pieces of the NVIDIA-Korea partnership.
- Whether Korean-language agents built on OpenShell reach production.
- Session recordings on NVIDIA On-Demand for any technical detail missing from the recap.
FAQ
What was AI Day Seoul?
A two-day NVIDIA event at COEX in Seoul focused on physical AI and AI agents, with more than 2,000 attendees.
Which Korean companies were involved?
NVIDIA cited Doosan Robotics, Hyundai Motor Group, LG Electronics and SK Group; speakers also came from Coupang Intelligent Cloud, LG AI Research, NAVER Cloud, SK Telecom and Upstage.
Were any products announced?
No new products or contracts were detailed in NVIDIA’s recap; the focus was talks, workshops, a hackathon and startup pitches.
Companies: NVIDIA
More on NVIDIA
AI · Nov 5, 2026
NVIDIA’s four open Cosmos 3 Nano models let robot and driving teams simulate worlds that reactAI · Nov 16, 2026
Rubin HPC is Nvidia’s bet that science agents will be bottlenecked by simulation, not reasoningAI · Nov 9, 2026
Nvidia’s Korean startup bets span interconnect, open models, robot hands and document AIAI · Nov 8, 2026
Nvidia’s Dynamo recipes speed four Korean open models to production, tuned for B200s