The competition in artificial intelligence (AI) semiconductors, led by Samsung Electronics and SK Hynix, is expected to expand into the realms of 'AI factories' and the physical AI era. This perspective suggests that AI should be viewed as a new industrial infrastructure, moving beyond generative AI and large language models (LLMs).
Jung So-young, CEO of NVIDIA Korea, delivered a keynote speech on the topic of 'New Industrial Revolution driven by AI Factory' at the AI Ecosystem Innovation Forum, held on July 9 at The Plaza Hotel in Jung-gu, Seoul, to commemorate the launch of the ABC economic broadcasting channel by the Aju Media Group. He emphasized that "data centers are not just facilities that incur costs, but are transforming into AI factories that generate new revenue."
Jung, a graduate of Seoul National University’s Department of Electrical Engineering, has held positions at IBM and Oracle Korea before becoming the CEO of NVIDIA Korea. He identified five key layers essential for driving AI: energy, semiconductors, infrastructure, AI models, and services. He stated, "When these five elements are organically connected without bottlenecks, AI can create new value. Currently, the biggest bottlenecks in the spread of AI are power, semiconductors, and data center supply."
He proposed the concept of the 'AI factory' as the next evolution of AI. Unlike traditional factories that produce goods through capital and labor, AI factories utilize electricity and data to generate new intelligence and tokens. He noted, "While existing data centers have been heavily cost-oriented, AI factories are production facilities that generate new revenue on their own. As we transition from dozens to hundreds of megawatt data centers to gigawatt-level AI data centers, the design and operational methods must change completely."
Jung also introduced NVIDIA's latest supercomputer platform, 'Vera Rubin,' highlighting how the competition in AI supercomputing platforms, which integrate GPUs, CPUs, networks, and interconnects, is changing the industrial paradigm.
Looking ahead, he identified agentic AI and physical AI as the next stages of AI development. He remarked, "In the future, most enterprise software will transition to agentic AI, and we are already in an era where AI understands problems and designs solutions autonomously."
He stressed that physical AI is not limited to humanoid robotics. "Physical AI encompasses all physical systems in the real world, from autonomous driving and smart factories to intelligent CCTV systems. Ultimately, the key to manufacturing innovation will be physical AI," he said.
To achieve this, he argued that simulation technologies for validating production lines and equipment through digital twins must advance alongside vision AI technologies that analyze workers and equipment in actual manufacturing environments. NVIDIA is expanding its collaboration with manufacturers through its 'Omniverse' and 'Metropolis' platforms.
Jung concluded by stating, "Korea is one of the few countries with competitive strengths in semiconductors, manufacturing, IT, and AI. The collaborative investment environment in AI between the government and private sector is also rare globally." He added, "Korea is one of the most important partners for NVIDIA in creating the physical AI era."
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.