NVIDIA, South Korean operator SK Telecom and AI chip startup Rebellions announced on Monday (November 9, 2026) a memorandum of understanding to create a heterogeneous computing system aimed at South Korea's sovereign artificial intelligence infrastructure using NVLink Fusion.
What happened?
The agreement involves using NVIDIA's NVLink Fusion technology to connect Rebellions' next-generation neural processing units (NPUs), known as Rebel R200, to NVIDIA's rack-scale infrastructure. SK Telecom plans to deploy these systems in its AI data centers, integrating NVIDIA GPUs with Korean NPUs to improve performance and energy efficiency in inference tasks.
The collaboration is supported by South Korea's Ministry of Science and ICT and was formalized during the AI Day Seoul event organized by NVIDIA. The goal is to enable Korea to develop sovereign AI capabilities tailored to local language, industries and priorities, reducing exclusive reliance on foreign hardware.
Why does this matter?
NVLink Fusion allows custom chipmakers to integrate their processors directly into NVIDIA's open infrastructure platform using high-bandwidth, low-latency interconnects. This facilitates hybrid systems that combine the raw power of GPUs with the specialized efficiency of NPUs.
According to the official announcement, Rebellions will use NVIDIA's MGX rack architecture for scale system development. SK Telecom intends to apply this to its full-stack AI strategy, including proprietary models and public services under the “AI for All” project. R200 systems are expected to be ready for deployment around 2028.
What changes in practice?
In practice, the partnership may enable disaggregated inference: NVIDIA GPUs for the compute-intensive prefill stage and Rebellions NPUs for the more energy-efficient decode stage. This tends to improve throughput and reduce power consumption in AI data centers.
For the Korean ecosystem, it represents an opportunity to bring domestic chips into the global AI infrastructure market alongside a world leader like NVIDIA. Jensen Huang, NVIDIA CEO, highlighted that sovereign AI gives nations the power to turn language, culture and expertise into their own intelligence.
Image credit: SK Telecom / NVIDIA | Source: Official NVIDIA announcement and Yonhap
By Grok
