With Chip Scaling Hitting Its Limits, Semiconductor Performance Now Hinges on Packaging and Substrates

The semiconductor battleground is no longer just about shrinking line widths; it now hinges on how densely and reliably multiple chips can be interconnected. As Moore’s Law has slowed, back-end processing and substrates have stepped into the spotlight. With AI accelerators, HBM, and chiplets all scaling up at once, the world outside the chip has suddenly become the most expensive bottleneck. Compute has to get stronger, memory has to sit closer, and power and heat have to be managed with far greater precision. That is why package substrates are adding more layers, wiring is getting finer, and everything from bonding materials and surface-treatment chemicals to solder balls and bonding wires is being reassessed.

In particular, the bill of materials for a single AI server has changed. The larger AI semiconductor platforms such as NVIDIA become, the more demand rises in lockstep for high-end CCL, high-layer-count MLB, FC-BGA, memory module substrates, lead frames, and packaging chemicals. But this is where we need to stay clear-eyed. We have to distinguish between beneficiaries already showing up in earnings and future opportunities that are still being validated. Glass substrates are clearly a candidate to reshape the next phase, but they are not yet in mass production. Expectations and revenue are not the same thing.

Advanced Substrates — PCB & Package Substrates Lifted by AI (6 companies)

As AI semiconductors get bigger and more powerful, the first real bottleneck is not the chip itself but the substrate and interconnect structure supporting it. In advanced substrates, the first name that stands out is Doosan’s Electronics BG. The fact that it is the sole supplier of high-end copper clad laminate (CCL) for NVIDIA Blackwell compute trays alone makes it crystal clear where the market is putting its money right now. Further up the stack, ISU Petasys has the scarcity value of being Korea’s only high-multilayer MLB player. Supplying 20–30+ layer high-multilayer MLBs for AI accelerators used by NVIDIA, Google, and Microsoft is not just a story of rising demand—it is a story of barriers to entry.

In package substrates, FC-BGA is the key battleground. Daeduck Electronics is one of Korea’s three FC-BGA players, while Korea Circuit is riding the expansion cycle in AI memory substrates with a portfolio spanning FC-BGA, system semiconductor substrates, and PCBs. Simmtech has a clear presence in MSAP-based package substrates such as FC-CSP and SiP, as well as memory module substrates. Haesung DS is even more interesting. On top of being the global No. 1 in semiconductor lead frame etching, it is also positioned to benefit from DDR5 and HBM demand through package substrates. In the end, this group shares one common trait: the better the chips become, the more sophisticated the substrate must be—and that level of sophistication is not something just any company can manufacture.

Company	Core focus
Doosan (Electronics BG)	High-end copper clad laminate (CCL) — sole supplier for NVIDIA Blackwell compute trays (Doosan business division)
ISU Petasys	High-multilayer MLB (20–30+ layers) — for NVIDIA, Google, and Microsoft AI accelerators; Korea’s only high-multilayer player
Daeduck Electronics	FC-BGA semiconductor package substrates — one of Korea’s three FC-BGA players (No. 1 is Samsung Electro-Mechanics)
Korea Circuit	FC-BGA, system semiconductor substrates, and PCB — expanding in AI memory substrates
Simmtech	MSAP package substrates (FC-CSP, SiP) and memory module substrates
Haesung DS	Global No. 1 in semiconductor lead frame etching + package substrates (benefiting from DDR5 and HBM)

Glass Substrates & Substrate Raw Materials — Materials That Could Change the Next Cycle (5 companies)

Glass substrates are a potential game changer that could replace silicon interposers, but this is less a market you can quantify today than one being validated for the post-2026 era. SKC, through its subsidiary Absolics, is preparing semiconductor glass substrates at its Georgia plant in the U.S. and has a production base built through a joint venture with AMAT, but it is still in the mass-production validation stage. JNTC has also moved into the space with its own TGV glass substrate technology on top of its strengthened-glass capabilities, and is targeting mass production in 2026 while currently at the sample supply stage. The excitement is real. But for now, it is still just excitement. The moment people start talking as if glass substrates will solve everything, judgment actually gets clouded. Validation always takes longer than expected.

That is why, in this phase, you need to look not only at glass substrates themselves but also at the surrounding raw materials. YCCHEM supplies the three core materials for glass-substrate TGV applications—photoresist, developer, and stripper—and counts Samsung and SK as customers. Kukdo Chemical ranks among the global leaders in epoxy resin with roughly 18% market share worldwide and 65% domestically, supplying key raw materials for CCL and semiconductor encapsulation materials (EMC). Wonik QnC supplies consumable semiconductor process parts such as quartzware and ceramics to Samsung, SK, Lam Research, and TEL. Even if the next cycle opens up, what ultimately supports it is this kind of raw material and consumable supply base. Technology themes may be flashy, but production is always completed by materials.

Company	Core focus
SKC	Semiconductor glass substrates — subsidiary Absolics’ Georgia plant in the U.S. (JV with AMAT), currently in mass-production validation stage (2026~)
JNTC	Strengthened glass + entry into proprietary TGV glass substrates — targeting mass production in 2026 (currently at sample supply stage)
YCCHEM	Three core materials for glass substrates (TGV), including photoresist, developer, and stripper — customers include Samsung and SK
Kukdo Chemical	Global leader in epoxy resin (~18%) and 65% domestic share — key raw materials for CCL and semiconductor encapsulation materials (EMC)
Wonik QnC	Consumable semiconductor process parts such as quartzware and ceramics — Samsung, SK, Lam Research, TEL

Packaging Materials — Connecting and Encasing the Chip (5 companies)

Packaging materials may not grab the spotlight, but they are the final link that turns a chip into a real product—and the first variable that determines yield. MK Electron holds a clear position as the global No. 1 in semiconductor bonding wire market share, and it is expanding into solder balls and solder paste. Duksan Hi-Metal is the global leader in micro solder balls for flip-chip bumping and ranks No. 1–2 worldwide in solder balls overall. As HBM and high-performance packaging scale up, these joining materials must become smaller and more uniform. One tiny ball can decide package reliability. Sounds trivial? Not at all.

YMT is Korea’s largest supplier of semiconductor surface-treatment chemical materials, with 248 product types, and it broke Japan’s monopoly in ultra-thin copper foil for package substrates. This is not just another localization story—it is about who controls the materials needed to realize fine circuitry. INNOX Advanced Materials, through INNOSEM, has semiconductor packaging bonding and adhesive materials, while Nano Shinmaterials has built touchpoints on both the process and materials sides with CMP slurry, sputtering targets, and CNT conductive additive dispersions. Packaging is not assembly—it is the sum of the process. That is why these materials companies look less like simple parts suppliers and more like businesses selling yield and performance together.

Company	Core strength
MK Electron	Global No. 1 in semiconductor bonding wire market share + expanding into solder balls and solder paste
Duksan Hi-Metal	Global No. 1 in micro solder balls (for flip-chip bumping) · No. 1–2 worldwide in solder balls
YMT	Korea’s largest semiconductor surface-treatment chemical materials portfolio (248 types) + ultra-thin copper foil (for package substrates; broke Japan’s monopoly)
INNOX Advanced Materials	Semiconductor packaging bonding and adhesive materials (INNOSEM) + OLED encapsulation materials
Nano Shinmaterials	CMP slurry · sputtering targets · CNT conductive additive dispersions

What to Watch Going Forward

The real beneficiaries of the AI supercycle are, first and foremost, advanced substrates and packaging materials that are already showing up in revenue; glass substrates and next-generation architectures are still areas where investors are essentially buying time. The areas with relatively clear near-term visibility right now are high-end CCL, high-layer-count MLB, FC-BGA, memory- and HBM-related packaging substrates, bonding wire and solder balls, and surface-treatment chemical materials. As Nvidia and AI server investment continue to expand, these segments are highly likely to translate into numbers quickly. Glass substrates, by contrast, are clearly attractive, but it would be misleading to ignore the fact that they are still in the mass-production validation stage. Technology validation for a post-2026 opportunity and investment decisions based on this year’s earnings should be treated separately.

One more thing. This industry cannot be explained by exciting technology stories alone. Substrates and materials are cyclical businesses, and when the market turns down, utilization rates and selling prices tend to weaken together. China’s catch-up is also intense. So the key question is always the same: who has actually secured customer approval and mass-production quality? Who is proving it not with words, but with volume? In an era defined by the limits of miniaturization, the winners will not simply be the companies that make things smaller, but the ones that can connect increasingly complex chips with the greatest stability. That is exactly why, when looking at semiconductors now, you can no longer focus only on the wafer front-end.

※ This article organizes company names, core businesses, and market positions based on publicly available materials verified online, and some assessments reflect the author’s own views. It is not intended as a direct basis for investment decisions.

Written: July 2026. Ealexandro