Samsung and SK Focus on 'AI' as the Key to Solving Semiconductor Process Challenges

Samsung Electronics Introduces AI Technologies Needed for Semiconductor Processes at AI Forum
SK Hynix Focuses on Developing AI to Improve Process Yield Through Subsidiary
Significant Impact Expected from AI Adoption in Semiconductor Processes

Choi Chang-gyu, Head of the AI Research Center at Samsung Electronics' Advanced Institute of Technology, is introducing AI technology applicable to the eight major semiconductor processes. / Source: Samsung Electronics YouTube channel capture

[Asia Economy Reporter Kim Pyeonghwa] The semiconductor industry, including Samsung Electronics and SK Hynix, is focusing on using artificial intelligence (AI) to enhance productivity and yield (the ratio of good products) throughout the semiconductor manufacturing process. As semiconductor processes continue to advance at microscopic scales, AI is seen as a key tool to solve the technical challenges that arise. There is also an assessment that the economic ripple effects resulting from this technological development are significant.

According to the semiconductor industry on the 10th, Samsung Electronics held the 'Samsung AI Forum (SAIF) 2022' on the 8th at the InterContinental Seoul COEX. SAIF is a platform for technological exchange aimed at exploring the latest AI research trends and future innovation strategies. At this event, Samsung Electronics highlighted semiconductors as a major industrial sector where AI can be applied. The Samsung Advanced Institute of Technology presented AI application cases in semiconductor processes and introduced future challenges.

Choi Chang-gyu, head of the AI Research Center at Samsung Advanced Institute of Technology, introduced that AI utilization in semiconductor processes can be enhanced by using factory monitoring (Fab Monitoring) and yield prediction technologies. He explained that detecting quality issues in advanced processes can save production costs, and building a digital twin across all processes to calculate wafer quantities can improve process efficiency. A digital twin refers to a computer-generated virtual representation of real-world machines or equipment.

Han Jae-jun, a master researcher at Samsung Advanced Institute of Technology, emphasized that AI usage could increase as semiconductor processes become finer. Entering the 3-nanometer (nm; 1 nm = one-billionth of a meter) process era, the importance of metrology grows beyond simple inspection, making AI a viable alternative to human labor. However, he pointed out that currently, there is a lack of relevant data and technical challenges that need to be addressed. He also advised expanding AI applications through data generation via 3D metrology and computer vision.

Conceptual diagram of three solutions developed by Gauss Labs / Source=SK Hynix Newsroom

SK Hynix, one of the two major pillars of the domestic semiconductor industry, is also taking steps to introduce AI into semiconductor processes. Currently, engineers use data to perform necessary tasks at each process stage, but in the future, the goal is to increase yield by adopting technologies such as automation through AI. SK Hynix is collaborating with Gauss Labs, an industrial AI company, to achieve this. Gauss Labs is a startup based in Silicon Valley, USA, established in August 2020 with capital invested by SK Hynix.

SK Hynix believes that the economic ripple effects of integrating AI technology into the semiconductor industry will be substantial. According to SK Hynix, the semiconductor sector in the industrial AI market is expected to secure a value of up to 100 trillion won in the mid to long term. Another advantage is that AI technologies developed for semiconductor processes can be easily applied or expanded to other industries.

To this end, Gauss Labs is focusing on developing AI process monitoring solutions that include functions such as metrology based on image data, virtual measurement, and monitoring. This month, they plan to officially launch PANOPTES Virtual Metrology (PANOPTES VM), an AI solution that predicts process performance related to unmeasured wafers.