Glass substrates will likely become a big investment theme in 2027, especially if we remain in a bull cycle, where themes tend to get fast-tracked
I’ve already done some coverage, including an interview with LPK’s CEO
In case anyone wants to check it out
Show more
Glass substrates are arriving sooner than planned.
NVIDIA is pressing its substrate makers to speed up development, which could bring adoption to before 2028. BOE has 24 layer substrates over 100 by 100 mm already validated in house and being tested with real server CPUs and GPUs, with a mass production investment decision by 1H27. Samsung Electro Mechanics and LG Innotek are targeting 2027 to 2028.
Glass replaces the conventional core inside the package with a much flatter and more stable material, so AI chips can be packed denser and run cooler. Packaging is turning into the next capex fight after CoWoS.
-TrendForce
Show more
Some updates on bottleneck timelines:
- ABF substrates: 1 year
/ eg. Ajinomoto (2802) for ABF film, Ibiden, Unimicron, Nan Ya PCB, etc make the substrates
- HDI Boards: Over 6 months
/ Victory Giant (300476), Zhen Ding (4958), Unimicron, Compeq (2313), Meiko (6787)
- Multilayer boards: ~6 months
/ Victory Giant, WUS (002463), $TTMI, Gold Circuit Electronics (2368),
- CCL: Over 6 months (severe shortages of materials)
/ Elite Material (2383), Shengyi (600183), TUC (6274), ITEQ (6213), Nan Ya Plastics (1303), Resonac (4004)
Mitsui Kinzoku (5706) / HVLP copper foil. Nittobo / glass fiber/cloth
- MLCCs, chip resistors, tantalum capacitors, aluminum electrolytic capacitors: 15-20 weeks to 1 year+
MLCC: Murata (6981), Samsung Electro-Mechanics, TDK (6762), Taiyo Yuden (6976), Yageo (2327), Walsin (2492)
Chip Resistors: Yageo, Walsin, KOA (6999), Rohm (6963) and $VSH
Tantalum capacitors: Kyocera/AVX, Yageo/KEMET, $VSH again
Aluminum electrolytic capacitors: Nippon Chemi-Con (6997) Nichicon (6996) | Timelines sourced from Digitimes, did the honors of adding in related companies.
But it's slightly more nuanced since MLCCs and others mix. in commodity types vs. AI DCs.
Regardless, just some helpful mapping.
Show more