Seoul National University Tech Holdings Bets on AI Software That Could Reduce GPU Dependence [fn Market Watch]

[Financial News] Seoul National University Tech Holdings has bet on cross-platform software technology that can run regardless of the type of artificial intelligence (AI) semiconductor. The move is aimed at reducing dependence on a particular GPU ecosystem and gaining an early foothold in the growing demand for diversified hardware as physical AI expands into robotics and autonomous driving.
On the 28th, Cha In-hwan, CEO of Seoul National University Tech Holdings, said, "The importance of cross-platform AI software that supports a variety of chips is growing by the day as the global AI market seeks to reduce the cost burden caused by the dominance of specific semiconductor chips." He added, "Vistrata is a team that has fully mastered low-level GPU control technology, which has extremely high barriers to entry. We decided to invest because the company is highly likely to secure the core standard platform in the era of mass-produced physical AI."
Vistrata is a deep-tech startup founded by veteran engineers with more than a decade of industry experience. The company is regarded as having both GPU control and machine learning optimization capabilities. Park Han-gil, its CTO and a KAIST Ph.D. who researched GPU rendering at Pearl Abyss and Netmarble, and Kim Hyeong-gyu, its CBO and an embedded systems and AI expert from Samsung Techwin and the Seoul National University Precision Machinery Research Institute, serve as co-CEOs.
Vistrata co-CEOs Kim Hyeong-gyu and Park Han-gil said, "Following this investment, we will focus on developing our core product, the AI runtime layer, and accelerate validation through collaboration with domestic and overseas companies in robotics and autonomous driving." They added, "We will grow into a standard-solution company that can respond flexibly to diverse AI hardware ecosystems around the world."
Meanwhile, the investment banking industry believes that as the physical AI market expands, investment targets are also broadening from finished robotic products and AI semiconductors to the software infrastructure that connects them.
A venture capital industry source said, "As the number of AI hardware options increases, the value of the software layer that absorbs performance differences between chips and lowers development costs will inevitably grow as well." The source added, "Competition over runtime and optimization technologies will also intensify in the physical AI ecosystem going forward."

[email protected] Kang Gu-gwi Reporter