The story of glass substrates is shifting from "who announces first" to "who can deliver on time." Entering the second half of 2026, a series of closely timed but differently oriented messages brought this change to the forefront: At the end of July, Intel and Lens Technology advanced their collaboration on glass substrate packaging for the AI era; on August 11, Korean equipment supplier Avaco and its affiliated company AVAT EC launched a TGV (Through Glass Via) process verification pilot line; on August 12, industry chain sources revealed that Samsung Electro-Mechanics' reliability evaluation of customer samples has encountered bottlenecks, and the construction plans for its joint venture GlaSSEM might be postponed, with mass production potentially delayed until after 2028.
Within the same month, we saw deals signed, pilot lines built, and demands for re-verification. This is not contradictory. Glass substrates have moved past the phase of "material performance comparison" and entered the engineering validation phase centered on customer reliability certification. Progress is no longer determined by the theoretical advantages of glass over organic substrates, but rather by whether glass, after processes like drilling, copper filling, circuit formation, and chip attachment, can withstand reliability assessments such as thermal cycling and humidity testing while maintaining flatness, low warpage, and stable electrical conductivity. Specific test items and thresholds vary by customer specifications, but this process cannot be compressed by capital expenditure or market hype, which is the root cause of recent adjustments in the industry's timeline.
01
Samsung Electro-Mechanics' Delay Exposes More Than a Single Point of Failure
Samsung Electro-Mechanics' delay represents a nine-month timeline. Market-leaked supply chain information shows that the company communicated its intention to purchase equipment for the GlaSSEM production line to suppliers as early as November 2025, but the order placement was postponed from December 2025 to March 2026, then to June 2026, with no new milestones provided since. Although key suppliers for processes like deposition, etching, laser drilling, and inspection have been largely confirmed, orders have not been finalized.
This is not due to the company's reluctance to invest. In early July, GlaSSEM signed a final agreement with Toyal Fine Chemicals, a subsidiary of Japan's Sumitomo Chemical, with a registered capital of 482.1 billion KRW; Samsung Electro-Mechanics holds a 66% stake. The joint venture's public goal is to be established in 2026 and establish a supply system in the second half of the 2027 fiscal year; this is not the same concept as full-scale mass production. The specific equipment lead times, the detailed construction schedule at the Pyeongtaek plant, and the subsequent mass production timeline still depend on customer validation and equipment introduction progress.
The real bottleneck lies in the samples. Current industry chain information points to issues in the customer sample reliability verification phase; however, the company has not publicly disclosed specific test items, failure mechanisms, or customer identities. The generally understood subsequent path involves readjusting glass formulations and TGV drilling parameters, resubmitting samples, and completing the customer's full re-testing process. The key to glass substrates is no longer just "whether micro vias can be drilled," but whether the substrate remains crack-free and maintains performance after chip attachment and heating.
Samsung is not alone. SKC and its U.S. subsidiary Absolics have also adjusted their previous commercialization pace, with market expectations shifting to their completion of final reliability testing within 2026 and advancing to full production in 2027. SKC has confirmed that its glass substrate business is preparing for customer reliability evaluations, planning to produce related samples and reviewing projects under discussion with multiple clients; however, customer lists, final certifications, and mass production orders have not been officially disclosed by the company. The industry's constraints have shifted from laboratory feasibility to stable yield rates and customer certification; these two cannot be conflated.
02
The Repeatedly Revised Mass Production Calendar
The timetables provided by various parties are currently quite conservative. Absolics's latest schedule involves completing reliability testing in 2026 and advancing to full production in 2027, later than previous market expectations. Samsung Electro-Mechanics' GlaSSEM aims to establish a supply system in the second half of the 2027 fiscal year, with the specific mass production timeline still uncertain. Dai Nippon Printing's TGV pilot line is already built, targeting the establishment of the system required for full-scale mass production in the 2028 fiscal year. LG Innotek had previously targeted mass production for 2027-2028, but subsequent commercialization pace still depends on technological and demand progress. TSMC views 2026 as a critical window for CoWoS equipment and material validation, with subsequent trial production and mass production timelines still subject to different assessments. Lens Technology has disclosed the construction of a related pilot line, conducting sample trial production and customer technical testing, but batch production still requires follow-up agreements and validation support.
Industry research institutions judge that TGV glass substrates might not enter the mainstream of advanced packaging until after 2030; the current phase is still early-stage validation, with the expectation of complementing rather than completely replacing existing packaging solutions. Domestic institutions are relatively more positive about the industrialization pace, viewing 2027-2028 as the first window for industrial validation and estimating that glass core substrates may see small-volume shipments starting from 2027, accelerating in volume by 2028. These are institutional forecasts or estimates, not yet realized industry facts.
A timeline shift to the right does not equate to disappearing demand. Public information shows that AI infrastructure and high-performance computing are driving demand intensity for high-end IC packaging substrates, and manufacturers are simultaneously expanding production. The concurrent growth in demand for traditional substrates and the delay in glass substrate mass production indicate that the current greater challenge for glass substrates lies in the technical maturity on the supply side, not in whether downstream demand exists. The market does not need another "theoretically better" material, but rather a material system that can be delivered stably under conditions of large size, high layer count, and high-frequency interconnects.
03
Intel and Lens Technology
The cooperation between Intel and Lens Technology is worth examining in detail. According to Lens Technology's announcement, the two parties signed a Memorandum of Understanding (MoU) focusing on TGV advanced packaging: Lens is responsible for glass substrate via formation, high-precision laser processing, metallized via deposition, and multi-layer interconnect routing; Intel intends to provide illustrative architecture information, Design for Manufacturing (DFM) guidelines, benchmark testing, and verification methods. The emphasis of this division of labor is not on a single order, but on the spillover of the certification system. The design side's willingness to provide DFM rules and verification methods means the glass processor will enter the actual packaging architecture design loop earlier.
This partnership has its practical basis. Lens Technology has disclosed the construction of a related pilot line, conducting sample trial production and submitting samples to some potential clients; some clients have entered the technical testing phase after passing preliminary proof-of-concept. The market also circulates progress such as a pilot line with a monthly capacity of 3,000 panels, a zero ppm non-conduction rate for millions of vias, minimum via diameters below 10 microns, a 30,000-square-meter TGV dedicated facility, and joint validation of a 22-layer glass core substrate; until these data are further disclosed by the company, they should not be equated with confirmed mass production facts.
However, an MoU cannot be equated with orders. Lens's announcement has explicitly cautioned that whether a formal agreement will be signed subsequently is uncertain, and as of the announcement date, the TGV business has not yet had a substantial impact on operating performance. For the industry, this is more like a capability positioning move; financially speaking, there is still a distance to forming verifiable business increments.
04
Japan Leads in Validation, Chinese Panel Makers Accelerate Entry
The advancement paths of Japanese and Chinese enterprises help in understanding the divergence in this round of mass production calendars. Shinko Electric Industries showcased a 22-layer glass core substrate in July — with 11 layers of copper wiring stacked on each side of the glass. Its technological focus is not just on the layer count, but on using protective materials at the substrate edges to disperse thermal stress, suppressing SeWaRe defects (cracks and delamination that may occur inside glass during processing or thermal cycling) under thermal load. Dai Nippon Printing, in December 2025, started a TGV glass core substrate pilot line in Saitama, aiming to establish the system required for full-scale mass production in the 2028 fiscal year. A common point for both companies is that their public actions focus on engineering aspects like multi-layer wiring, thermal stress control, and pilot-line verification, rather than simply announcing mass production.
Chinese enterprises have a different entry path. Industry research institutions listed 18 Chinese listed companies participating in the glass core substrate industry chain in August, covering segments like special glass, TGV processing, metallization, and advanced packaging. BOE's public progress is more representative: Its glass-based packaging substrate pilot line achieved full automation equipment integration in the first half of 2026, with a designed capacity of 1,000 panels per month; the company has completed the development and sampling of large-size, high-layer-count glass core substrates, with some domestic clients entering the technical testing stage. It must be emphasized that this business has not yet entered mass production nor generated mass production revenue.
Validation on the equipment side is also advancing simultaneously. The TGV pilot line launched by Avaco and AVAT EC supports large-format glass of 515×510 mm and connects key processes like laser drilling, etching, seed layer deposition, and optical inspection. Its significance lies not in immediately forming scaled capacity, but in verifying the repeatability and process-to-process efficiency for large-format glass across multiple critical steps. The actions of Japanese material and substrate companies, Chinese panel makers, and Korean equipment suppliers ultimately point to the same thing: Whoever can first organize glass processing capabilities into replicable semiconductor manufacturing capabilities has a chance to navigate the customer certification cycle.
05
Schott, Corning Push Glass Towards Optical Interconnect
The layout of material suppliers shows that glass substrates do not only have the single value proposition of "replacing ABF." Schott has publicly disclosed glass substrate and advanced packaging related products, with application directions covering high-density TGV, advanced heterogeneous packaging, and high-performance interconnects. Its products emphasize different coefficients of thermal expansion, thicknesses, and electrical properties to meet the material requirements of high-performance packaging.
Corning's direction is even more representative. The market-watched "Glass Bridge" concept applies TGV to next-generation Co-Packaged Optics (CPO) designs, extending glass from a packaging substrate to an optical interconnect substrate. Corning and NVIDIA have signed a multi-year commercial and technical collaboration, with the official disclosure focusing on expanding U.S. manufacturing capacity for optical connectivity and fiber; this collaboration itself should not be directly interpreted as a glass substrate or Glass Bridge project. Previously, BOE and Corning signed a three-year MoU covering directions like glass-based packaging substrates and optical interconnects.
This layout changes the judgment regarding initial applications. NVIDIA has publicly introduced its CPO technology direction, and Broadcom has termed its second-generation Tomahawk 5-Bailly as the industry's first production-ready CPO solution. This indicates the CPO ecosystem is entering a more concrete engineering phase. Compared to the longer industrialization cycle for glass core substrates, glass optical interconnect substrates for CPO may form an earlier entry point for implementation. Intel also positions glass as a foundational technology capable of providing both electrical and optical integration on the same platform.
However, material barriers should not be underestimated. Glass material, TGV via formation, metal filling, low-loss routing, thermal expansion control, and large-format manufacturing all involve patents and process accumulation. Once the industry matures, patent licensing and process route selection may become new variables in competition.
06
Yield is the Only Hard Currency
Underlying all the timeline debates is the same engineering problem: TGV yield. TGV is the process of drilling microscopic vertical vias through glass and filling them with copper to conduct electrical signals; each substrate requires thousands of such vias to be precisely aligned in a brittle material. Current glass core packaging yield is around 60% to 70%, far below the 80% to 90% of organic ABF substrates. The brittleness of glass means microscopic cracks during laser drilling can lead to the entire substrate being scrapped, making the cost 30% to 50% higher than ABF. MarketsandMarkets predicts the global glass substrate market will reach $8.4 billion by 2028, and Sigmaintell gives a similar figure of $8.5 billion, potentially reaching $27.6 billion by 2030 — but the premise for these numbers is that yield and certification issues achieve substantial breakthroughs within the next two years.
This round of intensive activity in the summer of 2026 essentially marks the first major test for glass substrates moving from the laboratory to the factory. The test is not over yet, but the passing line has been drawn.
This article is from the WeChat public account "Semiconductor Industry Perspective" (ID: ICViews), author: Jun Xi
