The context: a sovereign AI ecosystem
On November 9, Nvidia held its AI Day event in Seoul, where it announced a series of agreements with the South Korean government and several companies in the country, according to IT之家. The common thread linking the announcements is the idea of a sovereign artificial intelligence ecosystem, that is, built and managed at national level: the collaboration with the Ministry of Science and ICT (MSIT) covers AI semiconductors, data center infrastructure, training, startups, models, agents and security.
The agreement gives Korean NPU and XPU companies the possibility of exploring Nvidia's NVLink Fusion ecosystem, alongside training and certification programs, hackathons and technical seminars designed to increase GPU usage. This is an announced opening, not an already defined participation: the material explicitly speaks of exploration.
National models and infrastructure
Within the government's AI for All initiative, Nvidia will collaborate with four Korean foundation model developers to optimize and scale inference using Dynamo software: SK Telecom with A.X K2, LG AI Research with K-EXAONE 2.0, Upstage with Solar Open 2 and Motif Technologies with Motif-3. This is the area where the alliance takes on more concrete contours, because it ties inference software to already identified models.
On the infrastructure side, Nvidia announced a collaboration with SK Telecom and Rebellions, with MSIT support, for sovereign AI infrastructure based on NVLink Fusion, intended for Korean models and workloads. The involvement of Rebellions, a Korean chip company, signals the intention to integrate local suppliers into the Nvidia ecosystem.
From semiconductors to robotics
Samsung Electronics is expanding its collaboration with Nvidia to innovate in the semiconductor sector through accelerated computing and physical AI technologies, with activities covering materials research, process development and device engineering. Kakao Mobility, by contrast, aims to develop a Level 4 autonomous driving system leveraging Nvidia's compute and safety platforms, combining its own driving software with experience gained in Seoul's Gangnam district.
On the robotics front, Nvidia and LG Electronics said they successfully demonstrated the CLOiD robot in a real production environment, performing operations and manufacturing tasks guided by Isaac GR00T. Hyundai Motor Group is exploring Nvidia's NemoClaw blueprints and OpenShell runtime, while Korean AI security company AIM Intelligence is integrating its Starfort platform with OpenShell to control runtime access and detect malicious instructions and data leaks, with a possible pilot project in the public energy sector.
What remains to be verified
The overall picture shows a broad plan, but levels of maturity differ. Some announcements describe collaborations already launched and a demonstration already completed, as in LG's case; others speak of exploration, as with Hyundai and the NPU and XPU companies. No financial values, supply volumes, implementation timelines or performance metrics are indicated, and no statements from representatives of the parties are reported. The actual scope of the commitments, how to access the training programs and the fate of the announced pilot projects therefore remain to be clarified.
