Nvidia and Korean partners plan shared GPU and NPU inference systems

Nvidia, Rebellions and SK Telecom have announced a partnership to combine Korean-designed AI processors with Nvidia infrastructure for AI services. Rebellions' R200 systems are expected to be ready for deployment in 2028, so the announcement describes a development plan rather than an operating installation.

Rebellions plans to connect its next-generation R200 neural processing units through NVLink Fusion and use Nvidia's MGX rack architecture. SK Telecom intends to deploy the systems in its AI data centre and link them with its own models and public-facing services.

The companies aim to divide the stages of generating an AI response between different processor types. Their proposed approach uses Nvidia GPUs for prefill, which processes the input, and Rebellions NPUs for decode, which produces the response. Claimed throughput and energy-efficiency benefits remain goals to be tested.

Korea's science and technology ministry supports the collaboration. The plan seeks to connect domestic chip development with infrastructure for locally developed AI services. Successful delivery will depend on processor development, system integration and deployment, not simply the partnership announcement.

Dates: The partnership was announced in Seoul on 9 November 2026. Nvidia's page displays 8 November, while its Seoul dateline is 9 November. Deployment is planned for 2028.