[DigitalToday reporter Daegeon Seok] An AI chip designed in South Korea will be integrated into Nvidia rack-scale infrastructure. Nvidia said on Nov. 9 it is pursuing cooperation with Rebellions and SK Telecom to upgrade South Korea’s sovereign AI infrastructure based on NVLink Fusion. The three companies plan to combine Rebellions’ AI processors and SK Telecom’s data centre, model and service capabilities on Nvidia’s open infrastructure platform.

Rebellions will attach its Rivel R200 neural processing unit (NPU), a chip specialised for AI computation, to Nvidia’s platform using NVLink Fusion. NVLink Fusion is a technology that extends Nvidia’s high-bandwidth, low-latency scale-up networking to external custom chips. R200 units inside a rack will be linked with NVLink, and the system will be developed based on the MGX rack architecture.

Rivel R200, currently under development, targets low-latency, high-efficiency inference for advanced models such as mixture-of-experts (MoE) models with trillions of parameters. The completed system is set to be installed in SK Telecom’s AI data centres. Nvidia explained that GPU and NPU systems can share rack space, power supply and cooling design thanks to the MGX architecture. Rebellions will also use the MGX ecosystem’s ODM and components supply chain. The Rebellions R200 system is scheduled to be deployed from 2028.

The three companies will also develop a distributed inference approach that assigns the right processor to each stage of AI response generation. Under the configuration, Nvidia GPUs will handle prefill for input processing, while Rebellions NPUs will handle decode for output generation to improve throughput and power efficiency. SK Telecom plans to link this infrastructure with its own models including A.X K2, and with public services under the government’s Everyone’s AI project.

Rebellions CEO Seonghyeon Park (박성현) said the announcement is an important turning point for expanding South Korea’s sovereign AI ecosystem by combining Korean AI semiconductors and models with global technology. He said Rebellions will design next-generation AI infrastructure together with Nvidia and SKT, based on proprietary technology it has proven while growing as a South Korean AI chip company, and will work to make Korea-made technology a key axis driving global AI.

SK Telecom President and CEO Jaehun Jeong (정재헌) said the cooperation is meaningful because it demonstrates the potential of next-generation sovereign AI infrastructure by combining global AI infrastructure technology with domestic AI semiconductors. He said SK Telecom will strengthen full-stack sovereign AI competitiveness and contribute to the development of the domestic AI ecosystem by linking its AI data centres, its own AI models including A.X K2, and public-facing AI services developed through the Everyone’s AI project.

Nvidia founder and CEO Jensen Huang (젠슨 황) said South Korea has the talent, technology and industrial capabilities to emerge as a global leader in AI. He said NVLink Fusion will allow South Korean semiconductor companies to connect custom ASICs to Nvidia’s AI factory platform and run AI workloads by tightly integrating their own processors with Nvidia computing. He added that Nvidia is building a computing foundation with MSIT, Rebellions and SK Telecom so South Korea can produce its own intelligence and power the next industrial revolution.