NVIDIA, South Korean chip designer Rebellions and SK Telecom announced a collaboration on November 9, 2026, to build sovereign AI infrastructure for Korea, linking Korean-designed processors directly into NVIDIA's rack-scale platform. The agreement, supported by Korea's Ministry of Science and ICT, was unveiled in Seoul and brings together NVIDIA's open infrastructure, Rebellions' next-generation chips and SK Telecom's data centers, models and public-facing services.
At the center of the plan is NVLink Fusion, the technology NVIDIA uses to extend its platform to custom processors from other companies. Rebellions will use NVLink Fusion to connect its Rebel R200 neural processing units within a rack through NVIDIA's NVLink scale-up networking, and it will build its rack-scale systems around NVIDIA's MGX rack architecture and reference designs. That gives the Korean startup access to an established ecosystem of rack designs, component suppliers and original design manufacturers, rather than forcing it to build every part of a data center system on its own.
Rebellions is developing the R200 for low-latency, energy-efficient inference with frontier models, including multitrillion-parameter mixture-of-experts models. The partners also plan to explore disaggregated inference, in which different stages of generating an AI response are assigned to the processor best suited to each stage. Under the approach described in the announcement, NVIDIA graphics processors would handle the prefill stage and Rebellions processors would handle the decode stage, with the aim of improving inference throughput and power efficiency.
SK Telecom plans to deploy the resulting systems in its AI data center and connect them to its proprietary foundation models, including A.X K2, and to public-facing services developed under the government's AI for All project. The standardized MGX design means graphics processor and neural processor systems can share data center elements such as rack footprint, power delivery and cooling, which should simplify planning as Rebellions develops its processors and systems. The companies described the goal as full-stack sovereign AI, spanning infrastructure, proprietary models and services tailored to Korea's language, industries and priorities.
The government backing is explicit. Bae Kyung-hoon, Korea's Deputy Prime Minister and Minister of Science and Information and Communication Technologies, said the collaboration would let Korean-developed processors enter the computing infrastructure market alongside NVIDIA, and that the government would actively support it. NVIDIA chief executive Jensen Huang said sovereign AI gives nations the power to turn their language, culture and expertise into intelligence, while Rebellions cofounder and chief executive Sunghyun Park called the announcement a turning point for sovereign AI.
The announcement is a plan, not a finished product. Rebellions' R200 systems are expected to be ready for deployment in 2028, and the companies disclosed no investment amount, system size or purchase commitment. Even so, the structure matters. NVIDIA is increasingly willing to open its rack platform to outside silicon when doing so keeps its networking, racks and software at the center of national AI projects, and Korea is using that opening to make domestic chips a working part of its AI infrastructure rather than a side project.