Nvidia reveals direct investments in four Korean AI startups

Direct investment strategy

Nvidia unexpectedly announced its plans to invest in four Korean venture firms and startups during a side event at "AI Day Seoul." The disclosure occurred during the Nvidia AI Startup Partnership Forum held at the COEX convention center in Seoul. This move highlights a specific interest in the Korean technological landscape for artificial intelligence.

While Nvidia often uses its venture investment affiliate, NVentures, to fund early-stage companies, the reports indicate a shift toward direct investments from headquarters. This distinction suggests that the companies involved possess specific technologies that align closely with Nvidia's core corporate strategy. The investment process for most of these firms is currently ongoing rather than fully completed.

The companies identified for these investments represent diverse niches within the AI space. By selecting these specific startups, Nvidia is positioning itself to influence the development of foundational models and hardware interconnects. This strategy allows the company to secure early access to innovations that can be integrated into its broader hardware ecosystem.

Targeted Korean technologies

One of the primary targets for Nvidia's investment is Point2 Technology, which was founded by Bae Hyeon-min, the head of the KAIST Startup Institute. The company specializes in creating hardware that sends information between AI processors and storage racks within data centers. These components utilize radio frequency (RF) methods to enable fast communication.

In addition to hardware components, Nvidia is pursuing investments in software and model development. This includes Upstage and Motif Technologies, both of which are developers of large language models (LLMs). These models are the core engines behind many modern conversational and generative AI applications that companies use today.

The investment list also includes RealWorld, a developer of robot foundation models (RFMs). Robot foundation models provide the underlying intelligence that allows robots to perform complex tasks in various environments. By investing in these three distinct areas—hardware interconnects, LLMs, and robot models—Nvidia is creating a multi-layered presence in the Korean AI market.

Expanding the Korean ecosystem

Nvidia's move comes as part of a broader effort to support the Korean tech scene. The company has already supported more than 1,000 Korean companies through its "Nvidia Inception" program. This program is designed to help startups grow by providing access to resources, expertise, and networking opportunities.

The new investments represent a step beyond the standard support provided by the Inception program. By moving into direct equity or funding, Nvidia is taking a more active role in the growth of these specific entities. This transition from a support program to a direct investment model indicates a high level of confidence in the maturity of Korean AI startups.

For the Korean startups involved, these investments provide more than just capital. They offer a direct link to Nvidia's architecture and global distribution networks. This partnership is intended to help these firms scale their products to reach a global audience, moving them from local success to international competition.

Broader hardware alliances

The investment news coincides with other significant hardware partnerships involving Korean tech firms. Nvidia, SK Telecom, and Rebellions recently signed a memorandum of understanding regarding "NVLink Fusion." This platform is designed to extend Nvidia's high-speed GPU-to-GPU link capabilities to outside AI chips, specifically neural processing units (NPUs).

This marks the first formal cooperation between Nvidia and a Korean NPU company. Rebellions plans to connect its next-generation NPU, the R200, to Nvidia's high-speed connection technology. The goal is to create integrated systems where GPUs and NPUs can share the same rack space, power supply, and cooling facilities.

SK Telecom also plans to construct a joint GPU-NPU setup at its AI data center to verify performance and stability. These collaborations suggest that while Nvidia is investing in software and models, it is also working to ensure that Korean hardware can work seamlessly with its own architecture. This creates a comprehensive technological foundation for Korean-made chips to enter the global market.

What comes next for Korea

The trend of Nvidia's involvement suggests that Korea is becoming a critical hub for AI infrastructure and model development. By targeting both the hardware side (wireless interconnects) and the software side (LLMs and robot models), Nvidia is helping to build a full-stack AI environment in the region. This could lead to more Korean startups receiving similar direct investments in the coming years.

For the startups, the next phase will involve integrating their technologies into Nvidia's global ecosystem. Companies like Upstage and Motif Technologies will need to demonstrate how their models can scale alongside Nvidia's hardware. Success in these partnerships will determine how much influence Korean AI will have on the global market.

Industry observers will be watching to see which of the four startups successfully complete the investment process. The results of these deals may set a precedent for how other global chipmakers interact with the Korean startup scene. As the technology matures, the integration of Korean NPUs and models into global AI stacks will be a key metric of success.

Try it

Watch

Which of the four startups will complete the investment process first?

How the R200 NPU performs when integrated with Nvidia's NVLink Fusion technology.

Whether more Korean NPU companies will sign similar cooperation agreements with Nvidia.