SK Telecom is joining hands with Nvidia to target the global AI infrastructure market. It plans to build a "full-stack AI cloud" based on Nvidia's DSX platform and scale an AI work-focused "AI factory" to gigawatt capacity.

SKT said that at a meeting held in Taiwan on June 1, SK Group Chairman Chey Tae-won (최태원) and Nvidia Chief Executive Jensen Huang (젠슨 황) reviewed an AI infrastructure roadmap envisioned by the two companies and agreed to group-level cooperation. SKT will serve as a key implementer in the collaboration and push to build Asia's largest AI infrastructure.

◆First AI factory to start operating in 2027; expansion to gigawatt-scale infrastructure

The key is building an AI factory. An AI factory refers to an intelligent data center that continuously produces "tokens," a core unit of AI. It is designed as infrastructure specialized for AI training, inference and data processing.

The AI factory will begin its first operations in South Korea in 2027. SKT will expand the model to gigawatt-scale infrastructure and broaden the scope of its AI infrastructure business across Asia.

SKT will join Nvidia's global partner ecosystem program, the Nvidia Cloud Partner program. Through this, it will seek to secure the lowest token cost and the highest performance per watt, which determine competitiveness in the AI infrastructure market.

Cooperation with Nvidia is expected to become a core foundation for SKT's AI cloud business. SKT will support AI training and inference starting with Nvidia Blackwell graphics processing units and will also sequentially use the latest Nvidia Vera Rubin platform, which is scheduled to be supplied in the second half of this year.

Through the partnership, SKT will secure Nvidia AI infrastructure such as computing and dedicated software. Nvidia will leverage SK hynix's memory leadership and SKT's expertise in building and operating AI factories.

◆Shift from telecom carrier to AI cloud business operator

SKT will use the partnership as a springboard to rapidly grow its AI cloud business. As global big tech companies have recently been leasing large-scale AI infrastructure, AI cloud has emerged as a key pillar of the global AI infrastructure supply chain. SKT said its goal is to establish itself as a key partner providing the AI infrastructure and business network Nvidia needs across Asia.

While existing cooperation between SK and Nvidia was focused on semiconductors such as SK hynix's high-bandwidth memory, the scope of cooperation will be expanded to all areas of AI infrastructure, including building and operating AI factories. The companies will form a joint consultative body with Nvidia and also pursue joint research on next-generation AI factory architecture.

SKT and Nvidia will also cooperate in physical AI and robotics. At Nvidia's "GTC Taipei" keynote address held on June 1, large-scale digital twin technology built by SKT based on Nvidia Omniverse was introduced. SKT is applying the technology to SK hynix's semiconductor manufacturing process.

In robotics, it is advancing a robot simulation and training platform based on Nvidia's robotics platform, Nvidia Cosmos, and its humanoid AI model, Nvidia Isaac GR00T.

Chey said, "Based on a close partnership with Nvidia, we have secured full-stack AI infrastructure competitiveness," and added, "By jointly responding to issues including GPUs, memory and energy, we will be reborn as a leading AI cloud business operator driving the development of the AI ecosystem across Asia."

Huang said, "Telecommunications networks are evolving into national AI infrastructure," and added, "Telecommunications networks that connect people, companies and devices are now becoming the foundation of AI cloud."

He added, "Through Nvidia's DSX platform, SKT will be able to build a large-scale AI cloud and provide agent AI, enterprise AI and physical AI to leading companies and industries in South Korea and around the world."