LG Innotek is applying proprietary artificial intelligence technology to semiconductor glass-substrate manufacturing, aiming to improve both product reliability and production efficiency as global competition to commercialize the next-generation packaging material intensifies.
According to industry sources on Sept. 7, LG Innotek plans to integrate its in-house AI technology into the glass-substrate production process to maximize the efficiency of microcrack inspection. The company is reportedly collecting data on microcracks of varying shapes and sizes to improve the precision of its AI-powered inspection models.
“LG Innotek is working to dramatically enhance microcrack inspection performance with its own AI models,” an industry source familiar with the matter said. “It is training the system on a broad range of data, from defect-type classification to the effect of defects on the glass-substrate manufacturing process.”
Semiconductor glass substrates are regarded as a next-generation packaging platform for meeting demand for high-performance AI chips. Compared with conventional organic, or plastic-based, substrates, glass has lower thermal warpage and is better suited to ultra-fine circuit formation. Global chip companies and major technology firms—including Samsung Electronics, Intel, AMD and Broadcom—are considering their adoption.
Microcracks at the micrometer scale or smaller, however, remain one of the biggest obstacles to commercializing glass substrates. The defects can occur across complex manufacturing stages, including through-glass via (TGV) formation, metallization and substrate singulation. Because they are extremely small, they are difficult to detect completely with conventional inspection methods.
If microcracks are not detected early, the substrate can fracture or delaminate during subsequent processing, causing what the industry calls seware failure. Even a single microcrack can force manufacturers to discard an entire high-value finished substrate, resulting in significant yield losses.
“Because glass is a material with countless variables that can cause microcracks, controlling them is not easy,” another glass-substrate industry source said. “But introducing 100% inspection to identify every defect creates another dilemma: it can take too long, reducing the practicality of mass production.”
LG Innotek aims to overcome this bottleneck with AI. The company plans to combine high-resolution 3D precision inspection with its proprietary AI vision technology to improve detection capability while reducing analysis time. The objective is to secure both reliability and manufacturability by applying the company's accumulated artificial intelligence transformation, or AX, capabilities from other electronic-component manufacturing lines to glass-substrate production.
The initiative is viewed as a strategic move to gain an advantage in the emerging next-generation substrate market. SKC subsidiary Absolics and Samsung Electro-Mechanics are among the major Korean players pursuing glass-substrate technologies, while companies in Taiwan, Japan and China are also accelerating development efforts. In a field with an exceptionally high technical barrier, LG Innotek is positioning its proprietary AI capabilities as a key differentiator.
LG Innotek formally announced its plan to enter the semiconductor glass-substrate business at its annual shareholders' meeting in March 2024. It has since established a pilot-production line at its Gumi site in North Gyeongsang Province and is currently conducting reliability evaluations of prototype products.
Industry observers expect glass substrates to begin commercializing as early as around 2028. But the success of mass production is widely expected to depend on which company first masters microcrack-control technology—a critical factor in determining yield.
