SK Telecom is partnering with NVIDIA to establish a gigawatt-scale AI infrastructure. Their collaboration will extend beyond semiconductors to encompass full-stack operations in data center management.
On June 8, SK Telecom announced plans to pursue full-stack AI cloud collaboration based on NVIDIA's DSX platform. The two companies aim to gradually expand an AI factory specialized for AI tasks to a gigawatt scale.
The DSX platform is NVIDIA's comprehensive framework that defines the design, construction, and optimization methods for AI factories, covering everything from chips and systems to infrastructure software, facilities, and partner technologies.
This partnership was solidified during a meeting between SK Group Chairman Chey Tae-won and NVIDIA CEO Jensen Huang on June 1 in Taiwan. The AI factory is set to begin operations in Korea in 2027, with plans for SK Telecom to expand it throughout Asia.
SK Telecom will join the NVIDIA Cloud Partner (NCP) program to provide AI training and inference services based on NVIDIA's Blackwell GPU, with the Vera Rubin platform expected to be introduced in phases later this year. Key performance indicators will focus on minimizing token costs and maximizing performance per watt.
The scope of collaboration will also extend to research and development. SK Group and NVIDIA plan to form a consortium to jointly research next-generation computing architectures that optimize GPU and memory performance from the design stage. This marks a significant expansion of their existing semiconductor collaboration, which has primarily focused on SK Hynix's high-bandwidth memory (HBM).
Collaboration in the physical AI sector is also being formalized. SK Telecom is applying NVIDIA's Omniverse-based digital twin technology to SK Hynix's semiconductor manufacturing processes and is enhancing a robot simulation and training platform using NVIDIA's Cosmos and Isaac GR00T platforms.
Chey Tae-won stated, "With our close partnership with NVIDIA, we have established full-stack AI infrastructure capabilities that encompass everything from chips to data center operations. By addressing GPU, memory, and energy challenges together, we aim to become a leading AI cloud provider driving the development of the AI ecosystem across Asia."
Jensen Huang remarked, "Telecommunication networks are evolving into national AI infrastructures. The network that connects people, businesses, and devices is now the foundation of AI cloud. Through the NVIDIA DSX platform, SK Telecom will build a large-scale AI cloud, providing agent AI, enterprise AI, and physical AI to businesses and industries in Korea and around the world."
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.