SK Telecom is leveraging NVIDIA’s Omniverse platform to create a digital twin of SK hynix’s semiconductor production facilities, marking a significant step toward AI-driven smart manufacturing. The initiative seeks to replicate complex semiconductor fabrication environments in a highly detailed virtual space while enabling advanced simulation and optimization capabilities.
The company announced on June 1 that its manufacturing AI technologies were highlighted as a key industry collaboration case during NVIDIA’s AI conference, GTC Taipei, held in Taiwan. During the keynote presentation, NVIDIA showcased an ongoing project in which SK Telecom and SK hynix are jointly developing a digital twin model for semiconductor fabrication plants.
Digital twin technology creates a virtual replica of physical facilities and equipment, allowing manufacturers to test operational scenarios before implementing them in real-world production environments. The technology can be used to evaluate factory layout changes, equipment placement strategies, and process improvements, making it one of the most important AI applications in advanced manufacturing.
As part of its “Autonomous Factory 2030” strategy, SK hynix has been working with SK Telecom to accelerate the digital transformation of semiconductor manufacturing processes. The two companies completed proof-of-concept testing for the technology last year and plan to gradually expand deployment across production facilities.
At the core of the project is an “agentic digital twin modeling” technology that enables AI systems to understand manufacturing data and optimize virtual factory environments autonomously. SK Telecom has developed a platform using NVIDIA’s Agent Toolkit, allowing manufacturing equipment information and spatial data to be automatically analyzed, converted, and integrated into digital twin models.
The company said the technology significantly reduces the time required for 3D data conversion, scene generation, and performance optimization, tasks that traditionally require extensive manual effort. This is particularly important for semiconductor fabs, where thousands of interconnected tools, pipes, and support systems create highly complex environments that demand substantial computing resources during digital twin construction.
To address these challenges, SK Telecom has also worked on improving platform performance by enhancing GPU and memory utilization efficiency, helping to manage the large-scale computational workloads associated with semiconductor manufacturing simulations.
Industry experts regard semiconductor fabs as among the most challenging environments for digital twin implementation due to the extreme sensitivity of advanced manufacturing processes. Even minor operational variables can have a significant impact on production yields and process stability. As a result, leading semiconductor companies worldwide are increasingly investing in AI-powered simulation technologies to improve efficiency and manufacturing outcomes.
NVIDIA also highlighted the significance of the project. Mike Guyer, Director of Industrial Digital Twins at NVIDIA, noted that semiconductor fabrication facilities represent one of the most demanding manufacturing environments because of their vast 3D datasets and highly complex equipment structures.
“Semiconductor fabs combine massive amounts of 3D data with intricate facility configurations,” Guyer said. “SK Telecom has demonstrated the capability to apply Omniverse-based AI technologies to real industrial environments.”
