On June 1, 2026, Intel officially shipped the Xeon 6+ processor. This CPU, built on the 18A process node and integrating 288 processor cores, is the first to adopt a Glass Core thick glass substrate, making it the world's first commercial product utilizing a glass substrate solution. During the same period, TSMC (Taiwan Semiconductor Manufacturing Company) released the "CoWoS Glass Substrate Development Plan" to its supply chain, partnering with ABF substrate manufacturer Ibiden and panel manufacturer Innolux to jointly verify the feasibility of introducing glass substrates into advanced packaging; AMD unveiled its roadmap, planning to integrate glass substrates into its products by 2028; Samsung Electro-Mechanics has sent samples to Apple and Broadcom, with mass production targeted for 2027.
Leading manufacturers at home and abroad are taking intensive actions within the same time window, and glass substrates are moving from the laboratory to the production line.
01. Why Glass? Jetion
The core issue to be solved in advanced packaging is to widen and shorten the data channels between AI computing chips and HBM (High Bandwidth Memory). Taking NVIDIA Rubin as an example, a single package integrates 288GB of HBM4 (8 stacked dies), with a memory bandwidth of 22TB/s and an inter-chip NVLink6 bandwidth of 3.6TB/s. Supporting this bandwidth are thousands of sub-micron interconnects within the shared interposer—a technological path that is approaching the physical limits of silicon materials.
Image source: Collected and organized by MIR
Silicon interposers currently have three obvious shortcomings, which glass substrates can perfectly address one by one.
In terms of size, silicon interposers can only be manufactured on circular wafers, limiting their area—TSMC plans to expand the advanced packaging area by about 4 times in 4 years, and circular silicon wafers can no longer accommodate this. Glass, however, can be made into square panels of 510×515mm or even larger, free from the constraints of circular wafers. The material utilization rate can reach up to about 94%, and costs can be reduced by more than 20%.
In terms of warpage, packaging substrates expand when heated. The expansion of organic substrates is much more severe than that of silicon chips. As the packaging size increases, the substrate and chips are prone to deformation and misalignment, leading to low yields. The coefficient of thermal expansion (CTE) of glass is adjustable, allowing it to highly match that of silicon chips, making large-size packaging less prone to warpage.
In terms of cost and performance, silicon interposers have high manufacturing costs. Cutting circular wafers into square packaging units wastes a large amount of edge material, resulting in a material utilization rate of only about 58%. Glass is cheaper, with dielectric loss about 10 times lower than that of organic substrates, resulting in less high-frequency signal transmission loss and supporting higher data transmission rates. Additionally, the aspect ratio of Through Glass Vias (TGV) can reach 50:1, which is 5 times that of Through Silicon Vias (TSV), enabling finer and denser chip interconnects.
Simply put, the larger the silicon interposer, the harder it is to manufacture, whereas the larger the glass substrate, the more cost-effective it becomes. This is why it is regarded as the core material for next-generation advanced packaging.
02. How Far Has the Glass Substrate Supply Chain in Mainland China Come?
In the glass substrate supply chain, upstream raw glass sheets and core materials have long been monopolized by US and Japanese giants such as Corning, AGC, and NEG, making short-term breakthroughs difficult. The midstream equipment, as well as downstream glass substrate production, TGV processing, packaging and testing, and terminal applications, are the real main battlegrounds for domestic enterprises at present. The following focuses on these two segments to sort out the strategic layouts and core progress of various companies.
Glass Substrate Supply Chain in Mainland China
Image source: Collected and organized by MIR
2.1 Core Equipment: How Far Have Domestic Manufacturers Come?
Domestic substitution is moving from sample delivery to small-batch delivery.
In the field of laser drilling/cutting, DR Laser's TGV laser micro-hole equipment has achieved wafer-level and panel-level shipments, with small-batch repeat orders in 2026. This technology can also be applied in industries such as RDL, CPO, and new displays. De Long Laser's precision laser processing equipment covers the field of semiconductor advanced packaging, and TGV equipment is its key expansion direction in the IC (Integrated Circuit) packaging segment. Hans Laser possesses vertical integration capabilities from basic components and complete equipment to process solutions. The R&D of its glass substrate processing equipment has entered the small-batch stage, and phased results have been achieved in high-precision and high-efficiency laser drilling technology based on TGV and optoelectronic device substrate cutting machines.
In terms of electroplating/deposition equipment, Dongwei Technology's horizontal TGV electroplating line is the industry's first capable of batch horizontal transport of glass-based TGV filled-hole products, which can be used for 2.5D and 3D integrated semiconductor packaging. ACM Research (Shanghai) has a certain layout in the field of advanced packaging electroplating equipment. Sanfu Xinke is developing glass substrate electroplating equipment that is at the domestic leading level. It can optimize nozzle layout based on computational fluid dynamics simulation and use multispectral imaging technology for real-time detection of coating surface defects. Huicheng Vacuum has supply capabilities in the field of PVD coating equipment, providing support for TGV seed layer deposition.
In terms of lithography/exposure equipment, Smee's direct-write lithography equipment can be used for wafer-level and panel-level advanced packaging. The PLP series can meet the glass substrate's requirement for 2μm mass production resolution. The landing of PLP2000 orders marks the extension of direct-write lithography technology from wafer-level to panel-level applications.
In terms of inspection/cleaning equipment, enterprises such as Jingce Electronic, CAS Information, and Jetek are laying out equipment around the inspection and cleaning links in the glass substrate manufacturing process.
In terms of other supporting equipment, Hymson possesses full-chain independent R&D capabilities for laser + etching, with through-hole roundness reaching over 98%. As of May 2026, its TGV business has completed small-batch sample delivery. Enterprises such as Xinyichang, Union Laser, Keepprecision, and Hongtian Co. are also actively expanding in the field of glass substrate-related processing equipment.
2.2 Glass Substrate Production and TGV Processing: Who Has Mass Production Capabilities?
BOE is the participant with the clearest progress. Relying on its years of accumulated advantages in large-size glass densification processing, thin-film deposition, and large-scale integrated manufacturing, BOE invested 993 million RMB in 2024 to build a panel-level glass packaging substrate pilot line. In the first half of 2026, it achieved fully automated line throughput with a monthly capacity of 1,000 panels. It has fully completed the entire process of TGV hole opening, deep-hole copper filling, and build-up routing, and 20-layer large-size high-layer glass substrate samples have been sent to customers. In May 2026, BOE signed a memorandum of cooperation with Corning, focusing on key areas such as glass-based packaging substrates and foldable glass. The binding with upstream giants means that domestic substrates have obtained a ticket to enter the international supply chain.
Woge Optoelectronics is one of the few domestic enterprises with full-process capabilities for TGV glass substrates and has achieved mass production validation. The company's independently developed TGV technology adopts an ultrafast laser wet etching hybrid etching process, with a minimum hole diameter of 3μm and an aspect ratio of 150:1. The production line can process glass sizes of 510×515mm. Its subsidiary, Hubei Tongge Micro, has built an annual production capacity of 100,000 square meters for TGV glass substrates—this is currently the largest mass production capacity publicly disclosed in China.
Lens Technology publicly demonstrated over 3 million micrometer-level through-holes on 510×515mm glass with a 100% penetration rate in January 2026. The company is building a 30,000-square-meter TGV glass substrate factory, which is expected to be put into production by the end of 2026. Its independently developed TGV glass substrates have been sent to several leading domestic and overseas chip and packaging customers for sample testing.
2.3 Packaging, Testing, and Terminal Applications: How Far Is the Chinese Market from Commercial Application?
The downstream segment covers TGV packaging, display panel applications, and supporting materials and modules, which are the ultimate landing scenarios for the industrialization of glass substrates.
In terms of TGV packaging, leading domestic OSAT manufacturers such as JCET, TFME, WLCSP, and Huatian Technology are actively following up on the validation and introduction of glass substrate advanced packaging technologies. Yitian also has a layout in the packaging equipment field. In terms of display panel applications, enterprises such as Ledman, Leyard, Jufei Optoelectronics, Hongli Zhihui, and HC SemiTek are actively introducing glass substrates in the Mini/Micro LED display field, utilizing the high flatness and dimensional stability of glass substrates to improve display performance. In terms of supporting materials and modules, Debang Technology, Tianlu Technology, and Rokit Optics provide supporting support for the supply chain in fields such as packaging materials, optical modules, and photomasks.
So, how far is the Chinese market from commercial application? It can be judged from three dimensions: time, process, and cost.
From the perspective of the mass production timetable, 2026 is the industrial validation window, orders will be gradually released in 2027, and it will enter the volume ramp-up stage in 2028. Intel's TGV glass via substrate platform, aimed at long-term iteration, plans to start small-batch trial production in the second half of 2026. It is expected that the new generation of CPUs equipped with this solution will be launched from the end of 2026 to early 2027, achieving commercial readiness in 2027. TSMC's CoPoS panel will undergo trial production in 2027 and scale mass production in 2028, with the mass production of the complete CoPoS process integrating glass core substrates scheduled after 2030. Overseas leaders generally set the commercialization node at 2027-2028, and the Chinese market is highly likely to synchronize—BOE's pilot line throughput and Woge Optoelectronics' mass production line operation are precisely to seize this time window.
From the perspective of process maturity, the current mass production yield of glass substrates is about 60%, and yield ramp-up will be a long-term gradual process. Leading enterprises such as Intel have claimed that the yield is close to the commercialization level, but full substitution still takes time. TGV via processing, via filling reliability, and large-size yield breakthroughs remain key variables affecting the pace of mass production.
From the perspective of cost, the high cost of equipment replacement is the main obstacle. Panel-level electroplating, lithography, and other links require completely new production line investments. Compared with mature wafer-level processes, the upfront capital expenditure pressure is high. Only when the mass production scale is large enough and the yield is high enough can the economic viability of glass substrates substituting organic substrates be fully realized.
Comprehensive judgment: The display panel field (Mini/Micro LED) has relatively low cost sensitivity and high flatness requirements, and is expected to achieve large-scale introduction first. The commercialization in the field of AI computing chip packaging, however, must wait until the global mass production window opens in 2027-2028. For the domestic supply chain, the true commercialization turning point depends on the speed of yield ramp-up and the progress of downstream customer validation, which is expected to gradually emerge around 2028.
03. Future Development Trends
Given the deepening layouts by various parties and continuous breakthroughs in key processes, the industrialization of glass substrates is transitioning from the stage of "whether it can be made" to "whether it can be made well and cost-effectively." The following trends deserve close attention.
Trend 1: Continuous breakthroughs in process bottlenecks. The issues of glass brittleness and easy cracking have been systematically optimized through four dimensions: material selection, process optimization, structural reinforcement, and process control. The core competitive barriers in the industry have shifted from solving the basic problem of brittleness to high-level competition in high-precision via yield, multi-layer interconnection processes, and large-scale cost control.
Trend 2: Three technical routes advancing in parallel, with continuously expanding application scenarios. The implementation of glass substrates mainly unfolds along three main lines:
First is the panel-level packaging (CoPoS) solution: Compared with circular wafers, square glass has higher utilization rates, can be made in larger sizes, and has a cost advantage. It can also control thermal expansion and reduce warpage by adjusting the glass formula.
Second is the Glass Core Substrate (GCS) solution: Organic cores are prone to deformation when heated and have high high-frequency signal loss. Glass cores are more rigid, have lower loss, and have high-temperature warpage about 50% lower than organic substrates, with alignment accuracy improved by about 35%, making them more suitable for high-frequency, high-speed, and large-area chip interconnects. The 78×77mm, 10-2-10 stacked sample exhibited by Intel in 2026 follows this route.
Third is the CPO (Co-Packaged Optics) glass bridge solution: In the past, the docking of optical fibers and optical chips relied on external device alignment, resulting in high loss and limited channel spacing. Corning's Glass Bridge creates optical waveguides inside the glass, with a single connector supporting more than 24 optical channels and a minimum channel arrangement of 30μm. Under the same area, the integration density is increased by about 4 times compared to FAU (Fiber Array Unit), and the single-end face coupling loss is controlled within 1.5dB. After the volume ramp-up of 1.6T/3.2T ultra-high-speed optical modules, the market space for glass-based CPO substrates will rapidly open up.
Trend 3: Acceleration of the domestic substitution process. In the midstream TGV processing field, Woge Optoelectronics has achieved full-process capabilities and entered the mass production validation stage, and BOE's panel-level pilot line has completed throughput and sample delivery. On the equipment side, enterprises such as Hans Laser, DR Laser, and Hymson have achieved substantive breakthroughs in the field of TGV laser processing equipment, while Smee and Dongwei Technology are actively laying out in the lithography and electroplating links. Mainland China is gradually building a complete supply chain ecosystem from equipment to substrates, and from processing to applications. The localization of the raw glass sheet segment (with Kibing Group, Kaisheng Technology, Linuo Pharmaceutical Packaging, etc., accelerating their entry) will become the next direction for tackling key problems.
Conclusion
Currently, glass substrates have crossed the critical threshold from the laboratory to the mass production line. Intel's commercial shipments have proven technical feasibility, TSMC's CoPoS roadmap has locked in the mass production timetable, and the throughput of BOE's pilot line marks the substantive advancement of the domestic supply chain. The continuous explosion of AI computing power demand provides an irreversible driving force for this material revolution.
Looking ahead, the ramp-up of TGV via formation yield, the reliability validation of high-aspect-ratio copper filling, and the stability control of large-size panel-level processing will be the core variables determining whether glass substrates can move from "usable" to "highly usable." For the domestic supply chain, the progress in the midstream is already clearly visible. The real winning hand lies in: who can simultaneously solve the three problems of yield, cost, and customer validation within the 2027-2028 mass production window.
From organic to glass, from centimeter-level to micrometer-level, the material evolution of packaging substrates has just begun.