SK Group, Nvidia to build AI factory with 50,000 GPUs

Taipei [Taiwan], November 1 : SK Group said Friday that it is collaborating with Nvidia to build a large-scale artificial intelligence factory equipped with more than 50,000 of the US chipmaker's graphics processing units, according to Focus Taiwan.
Leveraging Nvidia's Omniverse -- a real-time 3D collaboration and simulation platform -- SK plans to establish a "manufacturing AI cloud" that will reinforce its competitiveness in the manufacturing sector while supporting the growth of related startup ecosystems.
The AI factory includes the manufacturing AI cloud and the AI data center project in Ulsan, forming an AI industrial cluster based on Nvidia GPUs. SK Group is advancing a hyperscale AI data center project in Ulsan targeting 100 megawatts by 2027, with plans to establish the site as a major Asia-Pacific AI hub.
SK hynix, the group's key affiliate and a leading global memory chip-maker, will implement digital twin technology across its semiconductor manufacturing lines. By creating virtual replicas of its production systems, SK hynix aims to improve yield rates, enable predictive maintenance and enhance operational efficiency -- all while significantly reducing costs.
The initiative also reflects Nvidia's broader vision to democratize access to high-performance GPU-based computing tools. The US tech giant said the AI platform will support startups in the manufacturing sector that currently lack access to such resources, helping drive innovation and lower barriers to entry.
SK hynix has long been one of Nvidia's closest partners. It is the exclusive supplier of HBM3 -- the fourth-generation high-bandwidth memory used in AI chips -- and accounts for roughly 75 percent of the global supply of the next-generation HBM3E.
"SK Group's collaboration with Nvidia is making AI the driving force for innovation across Korea's industrial landscape. Through this partnership, we expect to unlock new possibilities in scale, speed and precision for all sectors." SK Group Chair Chey Tae-won said. "Leveraging the Nvidia AI factory, SK Group will establish a next-generation infrastructure to power advances in memory, robotics, digital twins and intelligent AI agents."
"In the era of AI, a new kind of manufacturing plant has emerged: the AI factory," said Nvidia CEO Jensen Huang.
"SK Group is a vital memory technology partner, helping Nvidia create the world's most advanced GPU computing platforms that power global AI progress. We are delighted to partner with SK to build its AI foundation on Nvidia-accelerated computing and software, creating AI factories that will transform SK and energize Korea's AI ecosystem."
SK Telecom additionally signed a memorandum of understanding with Nvidia on the same day to jointly research and develop AI network technologies. The mobile carrier will collaborate with the US chipmaker to develop AI-RAN, a key technology for sixth-generation mobile communications.
AI-RAN enables the rapid, low-latency transmission of AI data generated by various devices across wireless networks. They will collaborate to validate, standardize and commercialize AI-RAN, with the intention to establish Korea as a global hub for AI-RAN technology verification.
Following SK Group's announcement of a strategic partnership with Nvidia, Chey held a separate meeting with Huang after the conclusion of the APEC CEO Summit in Gyeongju on the same day, during which the two reaffirmed their longstanding business friendship.
To make up for Chey's absence from the previous day's "chimaek" gathering -- a casual fried chicken and beer dinner -- in Seoul with other conglomerate leaders, Huang gifted him a bottle of Japanese Hakushu 25-year-old single malt whisky and Nvidia's DGX AI system, which is a high-performance computing platform that symbolizes the company's cutting-edge capabilities.
In return, Chey presented Huang with a semiconductor wafer produced by SK hynix and engraved with the message: "The world's best HBM4 product by SK hynix," highlighting the memory chip-maker's leadership in the high-bandwidth memory solutions essential for AI infrastructure.