NVIDIA, Rebellions and SK Telecom Plan AI Infrastructure in South Korea
NVIDIA, Rebellions and SK Telecom announced plans to integrate the upcoming Rebel R200 accelerator into NVIDIA infrastructure through NVLink Fusion. Deployment is expected in 2028.
NVIDIA, Rebellions and SK Telecom announced on November 9, 2026, that they are collaborating to build AI infrastructure in South Korea. According to the announcement, the Rebellions NVLink Fusion project will connect Rebellions’ upcoming Rebel R200 neural processing unit with the NVIDIA platform.
SK Telecom plans to use the resulting systems in its AI data centers and connect them with its own models and services. According to the announcement, the project is supported by South Korea’s Ministry of Science and ICT, known as MSIT. The financial value of the agreement, the number of systems and the specific data center locations were not disclosed.
Rebellions NVLink Fusion in the Joint Infrastructure Plan
Rebellions plans to connect its future Rebel R200 NPU to the NVIDIA platform using NVLink Fusion. It also plans to use NVIDIA MGX architecture in designing rack systems.
The announcement does not specify the exact rack configuration, the number of GPUs and NPUs to be used, or how individual workloads will be distributed between the chips in specific deployments.
One possible scenario, based on the project’s context, could be a combination of NVIDIA GPUs for the prefill phase and Rebellions NPUs for decode during inference. However, this is a hypothetical scenario, not a confirmed plan by the partners.
Deployment Planned for 2028
NVIDIA states that systems with the Rebellions R200 are expected to be ready for deployment in 2028. This is therefore an announcement of a technology and infrastructure objective, not confirmation of an available product or an already operating installation at an SK Telecom data center.
Rebellions has not yet disclosed the R200’s technical specifications, performance figures or energy efficiency. Independent benchmarks of the NVIDIA GPU and Rebellions NPU combination are also unavailable. The future deployment scale and binding orders from SK Telecom are likewise unknown.
The announcement uses the term sovereign AI, which the partners use to describe their strategic goal of building domestic AI infrastructure and connecting it with Korean models and services. For now, however, this is not an outcome that can be evaluated based on real-world operational deployment.
Integrating a Domestic Accelerator into the NVIDIA Platform
The collaboration presents a specific plan for how a South Korean AI accelerator manufacturer could enter rack infrastructure based on NVIDIA products. For Rebellions, integration through NVLink Fusion could provide a path to deployment in large AI data centers without building a completely separate hardware environment.
For SK Telecom, the key point is that the planned systems are intended to connect with its own AI models and services. The companies did not identify the specific products, model types or workloads these systems are expected to serve.
According to the announcement, MSIT is expected to support the collaboration. Details about the form or funding of support for this project have not yet been disclosed.
What to Watch
- the technical specifications, availability and production schedule of the Rebel R200 NPU,
- binding orders, deployment scale and locations of SK Telecom’s data centers,
- independent measurements of the NVIDIA GPU and Rebellions NPU combination,
- the specific form of support for the project from MSIT.
Sources
- NVIDIA Newsroom – Confirms the collaboration announcement, the planned use of NVLink Fusion and MGX, SK Telecom’s role, MSIT support and the goal of having the R200 ready in 2028.
- Ministry of Science and ICT, Republic of Korea – Documents the South Korean ministry’s broader policy for developing AI infrastructure and cooperation with NVIDIA; it is not independent confirmation of this specific agreement.
Verified and updated: 11/09/2026 06:19
