Nvidia, SK Telecom and Rebellions have agreed to connect a Korean-made artificial intelligence processor with Nvidia’s data center architecture, creating a route for a domestic chip to be tested within a globally used computing platform.

The agreement was signed at Nvidia AI Day in Seoul. Under the arrangement described by South Korea’s science ministry, Rebellions will link its neural processing unit with Nvidia graphics processing units through NVLink Fusion, while SK Telecom develops the artificial intelligence data center infrastructure needed to use the combined system.

NVLink Fusion is designed to let outside semiconductor companies integrate their accelerators into Nvidia’s rack-scale computing architecture. That could allow different types of processors to operate in one infrastructure, giving data centers more flexibility to balance computing performance and power efficiency. The evidence does not establish a commercial deployment schedule, customer commitments or measured performance gains.

The partnership comes as demand for inference grows. Inference is the stage in which a trained artificial intelligence model generates responses, and the ministry said dedicated chips can help address its power and operating-cost demands. Rebellions gains an opportunity to validate its technology in an international data center framework and seek overseas customers. The government plans to extend the approach to other Korean NPU companies and pursue links with platforms including AMD.