"AI Transformed Yields... The Future of Manufacturing Shown by Hynix" [AI Autonomous Manufacturing, Opening the Future]

Younghan Kim, CEO of Gauss Labs, Interview
"If AI Innovation Transforms Manufacturing, Korea Can Become a True AI Powerhouse"
"Reducing the Burden on SMEs with Subscription-Based Manufacturing AI"

"If AI innovation can occur in the industry we know best and in which we are most competitive, Korea could truly become a powerhouse in AI. I am determined to change the game."

Younghan Kim, CEO of Gauss Labs, emphasized this point in an interview with Asia Economy. He stated, "For Korea, a country centered on manufacturing, to break through in the AI era, we must focus on technologies that solve problems at industrial sites." He explained that if AI delivers tangible results in manufacturing, where Korea already holds a global competitive edge, this alone could further strengthen the nation's competitiveness on the world stage.

Gauss Labs is an industrial AI specialist company established in 2020. Its headquarters are currently located in Palo Alto, California, with additional offices in Seoul and Vancouver, Canada.

Younghan Kim, CEO of Gauss Labs, is explaining semiconductor manufacturing innovation using AI. SK Hynix

Since the end of 2022, AI technology has been applied to actual semiconductor processes, resulting in measurable outcomes. At SK Hynix's semiconductor factories, only a small fraction of the thousands of wafers constantly moving along the conveyor are selected for inspection with precision metrology equipment. Each wafer inspection used to take tens of minutes, and production continued under the assumption that most wafers would have no issues. However, since the introduction of Gauss Labs' AI, thousands of data points?including process conditions, equipment history, and sensor values for every wafer?are analyzed in real time by AI to predict quality. Through machine learning algorithms, the AI can forecast process results, enabling real-time control and monitoring of processes and equipment, as well as optimization of metrology.

The AI detected minute temperature changes of just 0.3 degrees, subtle equipment vibrations, and even recurring anomalies immediately after operator shift changes. The AI tracked variables that humans could easily miss. Kim stated, "Previously, we had to identify problems with very limited data, but now we can obtain complete data in real time, allowing us to detect and address issues much faster." In fact, after adopting this technology, process variability at Hynix decreased by 29%. This improvement was not just about saving one or two wafers; it enabled the recovery of thousands of additional finished chips. For Hynix, where high-value and high-margin products like server DRAM and high-bandwidth memory (HBM) account for a significant portion of production, this change could translate into additional profits worth tens of billions of won.

Kim explained that for AI to function as a practical tool on the manufacturing floor, "the environment is more important than the technology itself." Manufacturing data is complex in structure and changes in nature depending on conditions, making it difficult to apply general AI models. Gauss Labs adopted a structure that allows for repeated automatic learning and resetting, and designed tens of thousands of predictive models to operate simultaneously for each set of process conditions.

Kim diagnosed that closing the 'data infrastructure gap' is the top priority for spreading AI technology across Korea's manufacturing sector. While large corporations have the necessary infrastructure, most small and medium-sized enterprises do not. As a solution, he proposed establishing common data standards and introducing a 'subscription-based manufacturing AI service' model to lower the initial adoption burden. By providing AI technology through a software-as-a-service (SaaS) model, small and medium-sized enterprises can utilize AI without upfront investment, significantly lowering the barrier to entry. Kim emphasized, "For AI technology originating in manufacturing to flow throughout the industry, the government must support initial adoption by small and medium-sized enterprises and establish policy foundations that enable continuous data accumulation at industrial sites."