LG Electronics Strengthens AI Chiplet Business Below 7nm... Promising AI Chips in Home Appliances Sector

16th '8th AI Semiconductor Forum Breakfast Lecture'
Accumulating Customized On-Device AI Chiplet Manufacturing Capabilities
"Using All Foundry Dies from TSMC, Samsung, and Intel"

LG Electronics announced that it is advancing its sub-7nm (nanometer, one billionth of a meter) cutting-edge artificial intelligence (AI) semiconductor chiplet business in collaboration with foundry companies such as Samsung Electronics, TSMC, and Intel. Chiplets are a technology that separates each function of a chip into small pieces and then combines them into a single package, gaining attention as a solution to capacity and heat dissipation issues in high-performance computing (HPC). LG Electronics is enhancing its AI semiconductor business for home appliances and TVs and plans to expand its business, anticipating the greatest growth potential in AI semiconductors for the home appliance sector.

Jin-Kyung Kim, Executive Vice President and Head of the System on Chip (SoC) Center at LG Electronics, is giving a lecture on the theme of "AI Semiconductors for Smart Homes" at the 8th AI Semiconductor Forum Breakfast Lecture held on the 16th at the AC Hotel by Marriott Seoul Gangnam in Gangnam-gu, Seoul.
[Photo by Chae-Seok Moon]

On the 16th, Kim Jin-kyung, Executive Vice President and Head of LG Electronics’ System on Chip (SoC) Center, stated at the '8th AI Semiconductor Forum Breakfast Lecture' held at the AC Hotel by Marriott Seoul Gangnam in Gangnam-gu, Seoul, "We will actively utilize chiplets in next-generation products and use dies from TSMC, Samsung, and Intel," emphasizing this approach.

During his lecture on 'AI Semiconductors for Smart Homes,' Kim revealed that LG Electronics is focusing on chiplet development. With the recent surge in demand for data centers and HPC, chiplet technology, which effectively reduces power consumption and heat risks, is gaining traction. Previously, there was high demand for SoCs that integrated multiple chip functions into a single chip, but recently chiplets have emerged as an alternative.

He shared that since 2015, he has been intensively working to meet the business division's demand of whether LG Electronics can produce AI chips for home appliances and TVs better than Qualcomm, Intel, and NXP. While traditional cloud-centric AI requires enormous data input during the learning process, LG Electronics aims to reduce costs by focusing on on-device AI rather than cloud AI. On-device AI refers to technology that operates AI within devices without network connectivity.

After the lecture, Kim responded to reporters' questions about the process node size LG Electronics uses and its foundry partners, saying, "Although we cannot disclose exact figures, we are using very advanced processes, though not at the server-level (3~4nm)," and added, "Since chiplets are free from foundry constraints, we use dies from Samsung, Intel, and TSMC." When asked if the process is around 7nm or 5nm, he replied, "It is about that level."

He forecasted high utilization and growth potential for AI semiconductors in home appliances. When asked which sector has the highest demand for AI chips, he answered, "The diverse home appliance sector." Regarding whether home appliances are at a disadvantage compared to mobile application processors (APs) in terms of computing efficiency, he said, "Other companies buy design assets (IP) externally, but we are confident because we can customize IP optimized for customer applications."

On the 16th, key attendees of the "8th Artificial Intelligence Semiconductor Forum Breakfast Lecture" held at AC Hotel by Marriott Seoul Gangnam in Gangnam-gu, Seoul, are taking a commemorative photo.
From the third person on the left in the front row: Park Young-jun, Chief Technology Marketing Officer (CTMO) of Lifac; Yoo Hoe-jun, President of the Semiconductor Engineering Society (Professor of Electrical and Electronic Engineering at KAIST); Kim Jin-kyung, Head of LG Electronics System on Chip (SoC) Center (Executive Vice President); Yoo Hyun-kyu, Professor at Chungnam National University (Chairman of the Artificial Intelligence Semiconductor Forum).
[Photo by Moon Chae-seok]

Kim expressed confidence that LG Electronics is specialized in AI semiconductor chiplet business for home appliances and TVs and possesses superior capabilities in this field compared to rival home appliance companies including Samsung Electronics. He explained that the reason LG TVs perform better despite using similar display panels is due to superior chip performance. When asked if the secret to LG digital TVs’ better performance compared to competitors lies in SoC semiconductors, he replied, "Samsung and Sony also use our (LG Display’s) organic light-emitting diode (OLED) panels," and added, "Regarding image quality differentiation, the industry mainly mentions chip performance."

He also stressed the urgent need to nurture domestic fabless (design-specialized) companies. Currently, LG Electronics only produces semiconductors for premium products, while chips for general-purpose products are sourced from Taiwan’s Novatek and Chinese semiconductor companies. He said, "We only make chips for premium products, but we hope Korean companies handle general-purpose products," and questioned, "Domestic fabless companies have sufficient capability to make chips for products like home appliances and TVs, so why use Taiwanese and Chinese chips?"

Kim emphasized that China’s semiconductor rise should not be underestimated. Although progress is slow, he evaluated that the domestic industrial supply chain is strengthening. He predicted China’s semiconductor self-sufficiency rate will remain below 25% next year, while the Chinese government’s target is 70%. Kim stated, "Huawei’s Ascend 910C technology has many issues, requiring tens of thousands of technical support personnel due to trial and error, but despite U.S. sanctions, they will eventually produce AI semiconductors."