SK hynix said it plans to build an autonomous fab (semiconductor manufacturing facility) by 2030.
Do Seung-yong, SK hynix executive vice president (head of the DT institutional sector), said this on the 17th (local time) during a panel discussion at Nvidia's annual developer conference, "GTC 2026," held at the San Jose Convention Center in California.
Do said, "While demand for AI memory is rising rapidly, semiconductor production capacity is not keeping pace," adding that simply building new fabs will not solve the issue.
Do explained that the company will build an autonomous fab with the aim of simultaneously and timely pursuing new fab construction and productivity improvements at existing fabs.
Do added, "More sophisticated decision-making is needed, considering the balance among quality, expense, and speed," and "it is now difficult to meet these demands with human experience or rule-based automation."
Do then projected that the autonomous fab will dramatically shorten the transition time from design to mass production.
SK hynix is building the autonomous fab around three key pillars: "Operation AI," which serves as the brain; "Physical AI," which takes on the body; and "digital twin," which is responsible for the safe evolution of all elements.
Among these, Operation AI aims to automate not only routine tasks but also engineers' judgments, and SK hynix has cut processing times in areas such as maintenance and defect analysis by half through it.
Physical AI strengthens existing systems and expands automation into areas that still rely heavily on people.
Do noted in particular that the company used Nvidia's virtual-space library platform "Omniverse" to build this.
He explained that by using it, the company was able to simulate production flows, logistics movements, and process conditions before applying them, enabling learning and optimization without affecting actual production.