MOU signed to combine capabilities via “NVLink Fusion” First deployment in Korea at SKT’s AI data center Raising prospects for Korean technology to enter global markets
On the 9th, at “NVIDIA AI Day Seoul 2026” held at COEX in Gangnam-gu, Seoul, NVIDIA, SK Telecom, and Rebellions signed a memorandum of understanding (MOU) to connect domestic artificial intelligence (AI) semiconductors with NVIDIA GPUs within a single system, and then posed for a commemorative photograph. From left are Park Sung-hyun, CEO of Rebellions; Jeong Seok-geun, Chief Technology Officer (CTO) of SK Telecom; Lee Do-kyu, Director General for ICT Policy at the Ministry of Science and ICT; and Ashish Karandikar, Vice President of NVIDIA’s NVLink Fusion division. Provided by SK Telecom.
Domestic artificial intelligence (AI) semiconductors will now be able to operate together with NVIDIA’s graphics processing units (GPUs) in a single system. Attention is focusing on whether this will provide a springboard for Korean AI semiconductor companies to enter the global data center market based on NVIDIA’s technology.
According to the Ministry of Science and ICT on the 9th, SK Telecom and domestic AI semiconductor company Rebellions signed an MOU with NVIDIA at COEX in Seoul for cooperation on “NVLink Fusion.” The three parties will pursue technical collaboration to introduce into SK Telecom’s planned AI data center a system that combines NVIDIA GPUs with Rebellions’ next-generation neural processing unit (NPU), the “Rebel R200.”
NVLink Fusion is a platform that, based on NVIDIA’s high-speed semiconductor interconnect technology, enables external companies’ AI semiconductors to be integrated into NVIDIA’s systems. Using this platform, domestic NPUs can also operate in rack-scale AI computing systems equipped with NVIDIA GPUs. This is the first time NVIDIA has pursued such collaboration with a Korean NPU company.
The system being developed by the three companies will adopt a “disaggregated inference” method, in which the GPU and NPU share AI inference workloads. The GPU analyzes user queries, while the NPU, specialized for inference, generates the responses. “This collaboration is meaningful in that it validates the potential of a Korean-style sovereign AI infrastructure that can enhance inference performance and power efficiency by combining the strengths of NVIDIA GPUs and Rebellions NPUs,” said SK Telecom CTO Jeong Seok-geun.
This also represents an opportunity for domestic AI semiconductor companies to expand their overseas sales channels. “For Korean AI semiconductor companies seeking to enter overseas markets, the most important factor is securing a track record of operating together with NVIDIA or AMD products,” a Ministry of Science and ICT official said, adding, “Companies are saying that rather than participating in international exhibitions like CES or MWC 100 times, it is better to meet executives of global corporations directly and obtain opportunities for collaboration.”
Earlier in July, the government also signed an MOU with AMD on linking domestic NPUs with AMD products. Cooperation with NVIDIA will also be expanded to other domestic NPU companies. On the same day, the Ministry of Science and ICT and NVIDIA co-hosted an “AI Startup Partnership Forum” with the participation of 22 domestic AI startups to discuss cooperation measures in areas such as technology validation and commercialization.
Han Chae-yeon
AI-translated with ChatGPT. Provided as is; original Korean text prevails.
ⓒ dongA.com. All rights reserved. Reproduction, redistribution, or use for AI training prohibited.