[Edaily Reporter Lee So-hyun] #SKT is partnering with NVIDIA to build gigawatt (GW)-scale artificial intelligence (AI) infrastructure.

SK Telecom announced on the 8th that it will pursue an AI infrastructure business targeting the global market in partnership with NVIDIA.

The two companies will collaborate on a “full-stack AI cloud” that spans everything from chips to data center operations, based on the NVIDIA DSX platform. Their goal is to gradually expand the “AI Factory”—a data center specialized for AI workloads—to a gigawatt-scale.

NVIDIA DSX is a platform that defines the design, construction, and optimization methods for AI Factories across the entire full stack, including chips, systems, infrastructure software, facilities, and partner technologies.

Previously, SK Group Chairman Chey Tae-won and NVIDIA CEO Jensen Huang met in Taiwan on the 1st to review the AI infrastructure roadmap both companies had been envisioning. The two sides agreed to pursue group-level cooperation.

SK Telecom (SKT) will be the primary entity driving this collaboration. Based on its partnership with NVIDIA, SKT plans to pursue projects aimed at building Asia’s largest AI infrastructure.

An AI Factory is an intelligent data center that continuously produces “tokens”—the core units of AI—using power and data as raw materials. While existing data centers have focused on general-purpose computing and data storage, AI Factories are considered next-generation infrastructure optimized for AI operations such as training and inference.

SKT plans to launch its first AI Factory in Korea in 2027. Through this, the two companies aim to validate the governance and operational structure of their AI clouds, and subsequently scale up to GW-class infrastructure in phases before expanding across Asia.

SKT is also joining the “NVIDIA Cloud Partner” program, a global partner ecosystem that leverages NVIDIA’s high-performance AI infrastructure and software.

This collaboration is aimed at securing low token costs and high power efficiency, which are key factors in determining competitiveness in the AI infrastructure market. SKT plans to support AI training and inference starting with NVIDIA Blackwell GPUs and will sequentially adopt the latest NVIDIA Vera Rubin platform, scheduled for release in the second half of this year.

Implementing an AI factory requires full-stack AI capabilities that encompass not only computing but also networking, software, and data center operations.

Through this partnership, SKT will secure AI infrastructure, including NVIDIA’s computing and dedicated software, while NVIDIA will be able to leverage SK Hynix’s high-bandwidth memory (HBM) capabilities and SKT’s expertise in building and operating AI factories.

Based on this collaboration, SKT plans to rapidly expand its AI cloud business. Unlike general-purpose cloud services, the AI cloud is a business model that provides services specialized for AI tasks such as training, inference, and data processing.

Amid a trend where global Big Tech companies are leasing large-scale AI infrastructure, AI cloud is emerging as a core pillar of the global AI infrastructure supply chain.

SKT plans to establish itself as a key partner providing the AI infrastructure and business networks required by NVIDIA across Asia, beyond Korea, and to emerge as Asia’s leading AI cloud provider.

SK Group and NVIDIA will also pursue joint research to design next-generation AI factories and develop operational systems.

While the companies’ previous collaboration focused on the semiconductor sector, such as SK Hynix’s HBM, this agreement expands the scope of cooperation to the entire AI infrastructure domain, including the construction and operation of AI factories. The two companies plan to form a joint task force to research a new computing architecture that simultaneously enhances GPU and memory performance.

Cooperation in the fields of physical AI and robotics will also continue. During the keynote speech at NVIDIA’s “GTC Taipei” on the 1st, SK Telecom introduced its large-scale digital twin technology built on NVIDIA Omniverse. SK Telecom is currently applying this technology to SK Hynix’s semiconductor manufacturing processes.

In the field of robotics, they are advancing a robot simulation and training platform based on NVIDIA’s robotics platform “Cosmos” and the humanoid AI model “Isaac Groot.”

SK Group Chairman Chey Tae-won stated, “Based on our close partnership with NVIDIA, we have established competitive full-stack AI infrastructure capabilities spanning from chips to data center operations,” adding, “By jointly addressing challenges related to GPUs, memory, and energy, we will emerge as a leading AI cloud provider driving the development of the AI ecosystem across Asia.”

Jensen Huang, CEO of NVIDIA, remarked, “Communication networks are evolving into national AI infrastructure.” He added, “Through the NVIDIA DSX platform, SKT will be able to build a large-scale AI cloud and provide Agent AI, Enterprise AI, and Physical AI to businesses and industries in Korea and around the world.”