Nvidia, Rebellions and SK Telecom have announced plans at AI Day in Seoul, South Korea, for computing systems that would combine Nvidia’s infrastructure with Korean-designed chips and the telecom company’s data-center operations. The planned systems are expected to be ready for deployment in 2028, not available today.
The collaboration forms part of Nvidia’s AI Day at COEX in Seoul. A joint release dated November 8 sets out the plan, which Nvidia also featured in its November 9 event updates. Korea’s Ministry of Science and ICT is supporting the effort, which connects domestic chip development with AI infrastructure and services under the country’s AI for All initiative.
Rebellions plans to connect its next-generation Rebel R200 neural processing units, or NPUs, through Nvidia’s NVLink Fusion technology. The systems would use Nvidia’s MGX rack architecture, while SK Telecom plans to deploy them in its AI data center and connect them with its own models and public-facing services.
The plan divides the work of generating an AI response between different kinds of chips. Nvidia graphics processors would handle “prefill,” the initial processing of a user’s prompt. Rebellions processors would handle “decode,” the stage that generates the answer. That makes the Korean chips a planned complement to Nvidia GPUs, not a replacement for them.
For Rebellions, access to Nvidia’s rack design and supplier network would help turn a processor into a complete data-center system. MGX provides common design elements such as rack dimensions, power delivery and cooling for GPU and NPU systems. The companies say that shared foundation would support system development and infrastructure planning.
The broader aim is sovereign AI: computing infrastructure, models and services that countries can develop around their own language and priorities. SK Telecom plans to connect the proposed systems with its proprietary foundation models, including A.X K2, and services developed under AI for All. The collaboration therefore covers how chips, data-center operations and software would work together, rather than a chip announcement alone.
The companies said dividing inference work between GPUs and NPUs is intended to improve output capacity and energy efficiency. They did not disclose measured performance results, an investment amount or an order value. Those aims should not be read as results from a system already operating.
The key next step is development of Rebellions’ R200 systems for the expected 2028 deployment window. The announcement does not establish how many systems SK Telecom will install or a firm date for live service. Engineering, integration and deployment remain ahead.
JBizNews Desk | Seoul
© JBizNews.com. All rights reserved. Unauthorized reproduction or redistribution is strictly prohibited.
