SK Hynix to Build 'Autonomous Fab' by 2030... Creating an 'AI Brain' on Nvidia Platform

SK Hynix has announced its plan to build an autonomous fab (semiconductor manufacturing facility) by 2030.

Do Seungyong, Vice President of SK Hynix and Head of the DT Division, made this announcement during a panel discussion at Nvidia's annual developer conference 'GTC 2026' held on the 17th (local time) at the San Jose Convention Center in California, USA. He stated, "While demand for semiconductors is rapidly increasing, there are structural limitations that make it difficult for manufacturing to scale at the same pace."

Do Seungyong, Vice President of SK Hynix and Head of the DT Division, is participating in a panel discussion at Nvidia's annual developer conference, 'GTC 2026,' held on the 17th (local time) at the San Jose Convention Center in California, USA. Photo by Jinyoung Kim

He pointed out that as the proportion of high value-added and customized products, such as high bandwidth memory (HBM), increases, the complexity of fab operations is also rising. He diagnosed that the existing level of automation systems is no longer sufficient to keep up with the increasingly complex manufacturing environment.

Vice President Do added, "More sophisticated decision-making is required, considering the balance between quality, cost, and speed," and emphasized, "It is now difficult to meet these demands solely with human experience or rule-based automation."

He predicted that an autonomous fab would drastically shorten the transition speed from design to mass production. SK Hynix plans to build the autonomous fab based on three pillars: 'Operational AI' serving as the brain, 'Physical AI' as the body, and 'Digital Twin' responsible for the safe evolution of all elements.

Vice President Do explained, "Operational AI acts as the 'brain' of the factory, utilizing engineers' judgments and expertise in a data-driven manner to aid decision-making." He added, "As a result, we have reduced processing times by more than 50% in areas such as equipment maintenance and defect analysis."

Physical AI is intended to enhance existing systems and expand automation into areas still highly dependent on humans. This includes integrating AI into transport systems such as overhead hoist transport (OHT) for semiconductor wafers to make them intelligent. Vice President Do further explained, "By utilizing vision-based robots and autonomous mobile robots (AMRs), we aim to improve logistics efficiency and safety and expect to reduce parts inventory by approximately 30%."

These three pillars will be realized on a platform provided by Nvidia. Vice President Do stressed, "Based on Nvidia Omniverse, we are implementing the actual fab in a virtual space." He continued, "Through this, we can pre-validate production flows, material movements, and layouts, and we are performing simulation, AI training, and operational optimization without interrupting production."

This content was produced with the assistance of AI translation services.