LG Innotek announced plans to expand its business while unveiling semiconductor substrates for AI accelerators.
The substrates unveiled at the exhibition and the future mass-production targets are not the same thing.
The key point in this announcement is distinguishing the direction of large-substrate technology from the actual mass-production schedule.
3-Line Summary
1. LG Innotek unveiled an FC-BGA for AI applications.
2. The sample exceeds 100㎜ on one side.
3. 2027 and 2028 are mass-production targets.
Substrates for AI Accelerators Exceeding 100㎜ Were Displayed at the Exhibition
LG Innotek said it unveiled an FC-BGA for AI accelerators for the first time at KPCA Show 2026, the International Semiconductor Substrate and Advanced Packaging Industry Exhibition, held on 10 September at Songdo Convensia in Incheon. FC-BGA is a high-specification substrate that connects semiconductor chips and main boards.
The unveiled product is a large-area substrate exceeding 100㎜ on one side, and is a sample featuring ultra-fine circuits and high-layer-count technology. Newspim reported that the company also displayed an extra-large sample enlarged by 40% compared with existing products.
AI accelerators use high-performance semiconductors such as graphics processing units (GPUs) and neural processing units (NPUs). These products face the challenge of making internal substrate circuits denser while also increasing the substrate’s area. What LG Innotek unveiled this time can be viewed as the result of a product-development stage targeting that market.
As Substrates Grow Larger, Warpage and Wiring Become Challenges Together
Package substrates for AI semiconductors are evolving toward larger sizes and greater numbers of layers as the number of semiconductors and data pathways increases. Newspim reported a Samsung Electro-Mechanics official’s projection that high-performance substrates could expand from approximately 90㎜ and 20 layers to the 120㎜ class and 40 layers or more.
The problem is that substrates become more prone to warping as they grow larger and thicker. Simply making the core at the center of the substrate thicker can increase rigidity, but it may restrict the drilling of fine holes to connect the circuits above and below or the dense placement of circuits.
Since 2024, LG Innotek has promoted MLC (multi-layer core) technology as a key technology for large-area FC-BGAs. The technology combines multiple materials to reduce substrate warpage and improve fine-circuit and signal-transmission performance. The competition to make substrates larger is not merely about increasing their area; it is also a competition over how stably internal wiring can be implemented while controlling warpage.
2027 FC-BGA and 2028 Glass Substrates Are Targets
The point requiring the clearest distinction in this announcement is between the revealed samples and the mass-production plans. LG Innotek presented a goal of beginning mass production of high-value-added FC-BGAs for AI training and inference in 2027. It also set a 2028 mass-production target for next-generation glass substrates.
This is a business plan disclosed by the company; it does not mean that mass production at that point has already been finalized. Newspim reported that future competitiveness will depend on the design completeness of extra-large substrates and securing mass-production yields. Yield is an indicator of how stably and abundantly products meeting standards can be produced during the manufacturing process.
LG Innotek’s announcement shows a direction of expanding substrate technology built around smartphones and communications into high-value-added substrates for AI and servers. The distinction readers should make is clear. The unveiling of substrates measuring at least 100㎜ is a confirmed fact, while 2027 and 2028 are mass-production targets presented by the company.
References
- Newspim, LG Innotek Joins ‘2028 Glass Substrate’ Race…Direct Showdown with Samsung Electro-Mechanics
- Newspim, AI Substrate ‘Structural Competition’ Intensifies…Samsung Electro-Mechanics and LG Innotek Vie with MLC
Tags #LGInnotek #FCBGA #AISemiconductor #AIAccelerator #SemiconductorSubstrate #PackageSubstrate #MultiLayerCore #MLC #GlassSubstrate #KPCAShow #ArtificialIntelligenceSemiconductor #SemiconductorPackaging