Nvidia, SK Telecom and domestic AI chip startup Rebellions agree to work together on a heterogeneous AI computing system aimed at connecting Nvidia GPUs with Rebellions-designed NPUs. The cooperation is set out in a trilateral memorandum of understanding (MOU) announced in Seoul, with the system planned for use in building SK Telecom AI data centers.

The sources describe the effort as tying Rebellions’ NPUs into Nvidia’s rack-scale infrastructure using the U.S. chipmaker’s NVLink Fusion program. This approach is presented as enabling hyperscalers and custom chip designers to integrate their own AI processors with Nvidia hardware. Outlets also note the MOU is signed alongside AI Day Seoul, an Nvidia-hosted event in southern Seoul.

While coverage is largely consistent across reports, the Korea Times emphasizes the goal of accelerating AI innovation in Korea and highlights a quote from Nvidia’s Jesse Clayton. Yonhap’s versions frame the same announcement more straightforwardly, including multiple story variations that repeat the core details from the same briefing.