Glass Substrates: Samsung Electro-Mechanics, SK absolics and LG Innotek Face a Customer-Proof Test
Korea Supply Chain Brief · September 29, 2026
Korean securities analysts say the investment case for glass substrates is shifting from "who has the technology" to "who has paying customers and workable yields," according to a survey of the industry covering Samsung Electro-Mechanics, SK absolics and LG Innotek. Here is the state of each company's program, and why the global timeline for Intel and TSMC adoption matters for how soon Korean suppliers could see real revenue.
| Company | Role | Evidence |
|---|---|---|
| Samsung Electro-Mechanics KRX: 009150 | Running a glass substrate pilot line in Sejong; customer sample evaluation underway; mass production targeted for 2028 | Reported (Korean brokerage research, Sep 29, 2026) |
| SK absolics (SK Group) | Built a glass substrate production facility in Georgia, U.S.; customer certification and demand still pending | Reported |
| LG Innotek | Developing on a trial production line, scaling based on market demand | Reported |
Why the investment framework is changing
According to Korean brokerage research covered in the report, glass substrate investment logic has moved past whether a company has developed the technology and built capacity. Because mass-production schedules have slipped repeatedly and confirmed demand has not clearly materialized, the report says the market is now demanding concrete business proof: customer certification progress and actual signed orders, not roadmaps with target dates alone. One analyst, Park Hee-chul of Kyobo Securities, is quoted saying "a roadmap that only gives a target date is not enough — sharing certification-stage progress and yield trend data is the real task."
Why glass substrates exist as a category
Glass substrate is being developed to fix a specific reliability problem in today's organic (plastic) substrates: as AI chips add more high-bandwidth memory (HBM) and package multiple chips together, larger packages are more prone to warping and to reliability loss from mismatched thermal expansion between materials. Glass has high flatness and a thermal expansion rate closer to silicon than plastic substrates do, which suits large-area packaging — but it is more brittle and prone to microcracks during processing, so stable mass-production technique is the open technical question, not the material's theoretical advantage.
Company by company
- Samsung Electro-Mechanics: operating a pilot line at its Sejong plant and running customer sample evaluations now, targeting full mass production in 2028. The company is building out TGV (through-glass via) drilling, metal filling, and surface processing capability, expanding from its existing organic substrate business into glass.
- SK absolics: has secured a production facility in Georgia, in the United States. Customer certification and demand generation remain the open tasks, per the report.
- LG Innotek: developing technology on a trial production line, with a strategy of scaling capacity in line with market demand rather than building ahead of it.
The global piece: Intel and TSMC on different timelines
The report singles out Intel's commercial adoption as the market's key global signal to watch. Intel is pursuing a glass-core design that embeds a silicon bridge inside the glass core to connect chips, simplifying package structure while aiming to raise performance and power efficiency. Park is quoted saying Intel has a strong incentive to push glass substrate adoption aggressively to strengthen its foundry competitiveness, but cautions that "since the commercialization timeline and customers have not been disclosed, it needs to be confirmed whether expanded partnerships actually turn into mass production."
The report separately flags TSMC's CoPoS (panel-based packaging) as something to not conflate with glass substrate adoption: switching from round wafers to square panels as a production format is a different change from switching the substrate's core material to glass. TSMC's glass-core substrate commercialization is projected to come after CoPoS stabilizes, after 2030. Early demand for it is expected from CPO (co-packaged optics, which combines networking chips and optical components in one package, where glass's ability to carry both electrical and optical signal paths is an advantage) and from large-area advanced AI accelerators and custom ASICs.
What would move the story forward, per this report
Park's closing view: disclosing quantitative data on mass-production preconditions and whether cost parity with existing substrate technology has been reached will determine market confidence in the business case, and whether customer certification and firm orders actually materialize is what will drive any re-rating of these companies.
What is not disclosed
- Named customers for any of the three Korean companies' glass substrate programs.
- Specific yield rates or cost-parity data for any company's pilot or trial line.
- A confirmed commercial timeline or customer for Intel's glass-core packaging.
Quick FAQ
Which Korean companies are developing glass substrates?
Samsung Electro-Mechanics, SK absolics, and LG Innotek, each at a different stage — Samsung Electro-Mechanics is furthest along with a 2028 mass-production target from its Sejong pilot line.
Why does glass replace plastic substrates for AI chips?
Glass is flatter and has a thermal expansion rate closer to silicon, reducing warping and reliability problems in large AI chip packages with more HBM and multiple dies — though it is more brittle and harder to process without microcracks.
Is TSMC's CoPoS the same as adopting glass substrates?
No. CoPoS is a panel-based production format change; a glass-core substrate is a separate material change, and the report says TSMC's glass-core commercialization is expected only after 2030.
What's the actual risk analysts are flagging?
That capacity and technology development have outpaced confirmed customer demand and mass-production yield proof, making customer certification and order data the key thing to watch rather than roadmap announcements.
Source (Korean): 파이냸셜뉴스 (Financial News), Sep 29, 2026, reporter Choi Doo-sun. https://www.fnnews.com/news/202609291014193073
This brief is our own summary and analysis, not a translation. It is not investment advice. See our Disclaimer and evidence labels.
Stock Anatomy · Last checked: September 29, 2026