TLA

Nvidia, South Korean AI chip developer Rebellions and SK Telecom have announced a collaboration to build sovereign AI infrastructure in Korea, connecting locally designed neural processing units with Nvidia’s rack-scale platform and SK Telecom’s data centers, models and services.

The November 9 agreement, supported by Korea’s Ministry of Science and ICT, gives the country a concrete path for deploying domestic AI accelerators alongside Nvidia infrastructure. It also links the planned systems to SK Telecom’s public-facing services, including work under Korea’s AI for All program.

Connecting Korean NPUs to Nvidia infrastructure

Rebellions plans to use Nvidia NVLink Fusion to connect its next-generation Rebel R200 neural processing units within racks. The company will also use Nvidia’s MGX rack architecture and reference designs while developing the systems.

NVLink Fusion is designed to let custom processors connect to Nvidia’s scale-up networking and infrastructure ecosystem. Under the planned architecture, Nvidia GPUs would handle the prefill stage of AI inference, while Rebellions NPUs would perform the decode stage. The companies said this division could improve throughput and power efficiency, although they did not disclose benchmark results or a testing schedule.

Rebellions is developing the R200 for low-latency inference on frontier models. The company said its chip will support large mixture-of-experts models, but the announcement did not provide final specifications, pricing or manufacturing details.

SK Telecom plans to deploy the systems in its AI data center and connect them with its proprietary foundation models, including A.X K2. Its role covers data center operations, models and public services, while Nvidia supplies interconnect and rack-platform technology and Rebellions contributes the locally designed accelerator.

Deployment targeted for 2028

The collaboration is part of South Korea’s broader effort to build domestic AI capabilities spanning chips, infrastructure and services. The ministry said it will support cooperation between Nvidia and Korean NPU companies, beginning with Rebellions and SK Telecom, and also work with Nvidia on physical AI and GPU supply.

Rebellions’ R200 systems are expected to be ready for deployment in 2028. SK Telecom said it intends to connect the resulting infrastructure with its models and services as part of its full-stack sovereign AI strategy.

South Korea picks SK Telecom, Kakao and KT for free national AI service