Nvidia invests in Point2, plans stakes in 3 more Korean startups
The stakes come as Nvidia widens its work with Seoul's science ministry, from chip interconnects and inference software to robots and self-driving cars.
Summary
In brief
- Nvidia invested in Point2 and plans to invest in Motif Technologies, RLWRLD and Upstage
- Korean chip designers are to explore joining Nvidia's NVLink Fusion ecosystem
- Nvidia released Dynamo deployment recipes for 4 Korean open models
Nvidia said on Monday it had invested in Point2 Technology, a Korean startup that builds chip interconnects, and plans further investments in Motif Technologies, RLWRLD and Upstage.
The stakes are part of an expanded collaboration with South Korea's Ministry of Science and ICT that, according to the company's post, spans chips, data centers, talent, startups, models, agents and security. It also lays the groundwork for Korean chip designers to explore joining NVLink Fusion, Nvidia's program for linking other companies' processors to its systems.
Nvidia announced the investments at a startup forum it hosted with the ministry at COEX, where Vice Minister of Science and ICT Ryu Je-myung joined Nvidia leaders to meet more than 20 Korean startups, Nvidia's announcement said. Point2 builds radio-frequency interconnects that move data between chips and racks, an alternative to copper and optical links.
Motif, a Seoul-based lab, develops foundation models for agents, including the open-weight Motif-3. RLWRLD builds robotics models for tasks that need five-finger dexterity, and Upstage makes Solar language models and document tools. Nvidia said it counts more than 1,000 Korea-based members of Inception, its startup program.
The company also released deployment recipes for Dynamo, its open-source software for running models across many GPUs, covering SK Telecom's A.X K2, LG AI Research's K-EXAONE 2.0, Upstage's Solar Open 2 and Motif-3. Nvidia said the recipes, tuned for its B200 chips, deliver up to 2x performance out of the box. The figure has not been independently verified.
Under Korea's AI for All initiative, Nvidia will work with three selected consortia and four foundation model builders on inference, the running of trained models, and with SK Telecom and chipmaker Rebellions on NVLink Fusion infrastructure.
In a separate post, Nvidia said Samsung Electronics cut one chip simulation from about 18 hours on a 32-core CPU to 41 seconds with its tools. Kakao Mobility plans to build a level 4 autonomous driving system, which needs no driver in set areas, on Nvidia's Hyperion platform, starting with Kia vehicles.
On agents, Hyundai Motor Group is exploring Nvidia's OpenShell runtime, a sandbox that controls what software agents can access, and LG AI Research is opening API access to its K-EXAONE-2.5 model, Nvidia said.
The announcements follow President Lee Jae Myung's record 2027 budget proposal on November 2, which nearly doubles spending on chips and AI. In October 2025, Nvidia said it would supply more than 260,000 of its GPUs to the Korean government and companies.