Glass substrates, a dark horse in 2026

The industry's attitude towards glass substrates has done a 180-degree turn.

In the middle of last year, the media was still reporting rumors that Intel had abandoned glass substrate technology, and the market was still arguing about the commercialization prospects of glass substrates.

Companies such as Absolics and LG Innotek were still testing the waters on the production line, and most players in the ecosystem were in a wait-and-see state.

However, just half a year later, the goal of "mass-producing glass substrates in 2026" began to appear in the roadmaps of manufacturers, and research institutions also gave market growth forecasts that arouse people's imagination.

The background of glass substrate technology is that as generative AI training models evolve towards the scale of trillions of parameters, the physical performance of computing power infrastructure is facing severe bottlenecks. Against the background of the slowdown of Moore's Law, advanced packaging technology is no longer just simple chip assembly but a key path to improving the performance of semiconductor systems.

Currently, traditional organic substrates are gradually approaching their physical limits in terms of heat dissipation efficiency, large-size processing stability, and interconnection density. In contrast, glass substrates have significant physical advantages in terms of flatness, thermal stability, insulation performance, and interconnection density.

Where has the development of glass substrates reached today? Will this technology become a dark horse in the semiconductor industry in 2026?

Let's start with the industry report.

Market Size and Trend Forecast

According to data released by several authoritative market analysis institutions at the end of 2025, the glass substrate industry is experiencing a critical turning point from technology verification to early mass production.

The market generally predicts that 2026 will be the node for glass substrates to enter small-scale commercial shipments, and the period from 2028 to 2030 will be a period of rapid growth. A report released by Yole Group in November 2025 pointed out that from 2025 to 2030, the compound annual growth rate of semiconductor glass wafer shipments will exceed 10%.

It is worth noting that in the specific segment of memory (HBM) and logic chip packaging, the compound annual growth rate of the demand for glass materials is expected to be as high as 33%. This data directly reflects the urgent demand for high-density interconnection technology in the high-performance computing field. This demand does not come from the replacement of the existing market but from the explosion of the incremental market.

In terms of value distribution, data from a report by MarketsandMarkets in October 2025 shows that the global semiconductor glass substrate market size is expected to grow from $7.1 billion in 2023 to $8.4 billion in 2028.

Although the absolute growth of the overall scale seems stable, the incremental structure has undergone fundamental changes. The new added value is mainly concentrated in the high-end flip-chip ball grid array (FC-BGA) and advanced packaging fields. This means that the value of each unit product has increased significantly. Glass substrates will no longer be cheap carriers but core components mainly carrying high-value AI accelerators and server chips.

In terms of application scenarios, the first batch of commercial applications will be highly concentrated in ultra-large-scale data centers. For top-level AI training chips of leading customers such as NVIDIA, AMD, and AWS, they are less sensitive to costs but have extremely high requirements for computing power density and energy efficiency ratio.

The industry predicts that with the release of production capacity by leading manufacturers such as Absolics, it is expected that by 2030, glass substrates will gradually replace organic substrates in the high-end HPC market and become the standard configuration for the integration of trillions of transistors.

Global Industrial Competition Pattern

Facing the window period of technological iteration, the global semiconductor industry chain accelerated its production capacity layout in the fourth quarter of 2025. Major enterprises in South Korea, Japan, the United States, and China have all updated their mass production schedules, forming a competitive pattern of different technological routes and business models.

South Korea

The South Korean semiconductor industry has adopted an extremely aggressive vertical integration strategy, trying to build barriers during the window period in 2026 through internal cooperation within the consortium. Absolics, a subsidiary of SKC, has completed the installation of major equipment at its glass substrate factory in Covington, Georgia, USA, and has started providing mass-production samples to customers including AMD, entering the certification stage. Market sources said that the company plans to gradually introduce mass production from 2025 - 2026, first running through the yield rate and customer certification with limited production capacity to seize the first-mover advantage in the commercialization of glass substrates.

Different from SKC's single-point breakthrough, Samsung Group has demonstrated strong supply chain integration capabilities. In November 2025, Samsung Electro-Mechanics (SEMCO) established a joint venture with Sumitomo Chemical of Japan to specifically produce core glass core materials, ensuring supply from the source. At the same time, the pilot production line of Samsung Electro-Mechanics at its Sejong factory has been put into operation. Downstream, Samsung Electronics is actively testing the application of glass substrates in the packaging of next-generation HBM4 memory, trying to optimize the heat dissipation performance through the glass interposer.

LG Innotek has taken a differentiated route. In December 2025, the company upgraded its glass substrate working group to an independent division and announced the construction of a pilot line at its Gumi factory. Its R & D focus is not limited to electrical interconnection but to overcome the optical and electrical hybrid transmission technology, aiming to enter the future co-packaged optics (CPO) market. In addition, South Korean equipment manufacturers are also closely collaborating to accelerate the construction of the local ecosystem. Philoptics has delivered core laser equipment for large glass cutting to leading customers; Hanwha Precision Machinery has optimized its packaging equipment to deal with the fragility of glass substrates during handling, leveraging its dual accumulation in the display and semiconductor fields.

Japan

Japanese enterprises are using their accumulation in materials science and display panel equipment fields to implement a differentiated competition strategy for large-panel packaging. Rapidus included an independent advanced packaging production line in its expansion plan for the Hokkaido factory in December 2025. Its strategy is to directly use 600mm×600mm rectangular glass panels for packaging. This route aims to take advantage of Japan's accumulation in lithography machines and panel manufacturing to reduce unit costs by increasing the single exposure area. Rapidus has reached in-depth cooperation with IBM and DNP and plans to start mass production in 2028.

Among material manufacturers, DNP has built a test production line for TGV glass substrates at its Kuki factory in Saitama Prefecture for mass production verification, planning to supply samples in early 2026 with the goal of achieving large-scale production in 2028. At the same time, the material giant Resonac is accelerating the R & D of next-generation packaging materials suitable for glass substrates, focusing on overcoming the problems of interface bonding and stress control of glass materials in heterogeneous integration to establish its dominant position at the material end.

United States and China

Intel confirmed to the media last September that it would proceed with its commercialization plan for semiconductor glass substrates as originally planned, refuting reports that it might withdraw from this business due to operational challenges. The company reiterated that its commitment to developing glass substrates as a key technology for next-generation semiconductor manufacturing has not changed. Intel's semiconductor glass substrate development project still aligns with the technology roadmap formulated in 2023, with no changes to its schedule or goals.

At the IMAPS 2025 exhibition, Intel once again emphasized that glass substrates solve the key miniaturization challenges in advanced packaging, enabling finer feature miniaturization, larger packaging sizes, and enhanced high-speed I/O performance, strengthening the industry's confidence in its technology reserves. According to the plan, Intel's glass substrate products are expected to be widely used between 2026 - 2030.

TSMC is accelerating the development of glass-based panel-level fan-out packaging (FOPLP). Sources said that TSMC plans to establish a mini production line in 2026, initially using the 300mm specification and later transitioning to the large-panel process. TSMC is closely collaborating with Corning's Taiwan factory to jointly develop special glass carriers suitable for the CoWoS process to ensure the continuity of its advanced packaging ecosystem.

BOE established semiconductor glass substrates as a core strategy at its press conference in December 2025, planning to use the depreciation advantage of panel production lines and glass processing capabilities to achieve mass production of high-aspect-ratio products in 2027. Woguang Optoelectronics's Tongge Micro achieved small-scale shipments to overseas customers in the second half of 2025, mainly for microfluidics and radio frequency applications. In terms of equipment, manufacturers such as Han's Laser have started delivering domestic TGV laser drilling equipment, gradually breaking the overseas monopoly.

Packaging and testing and manufacturing enterprises are also accelerating the transformation of their technology reserves. Tongfu Microelectronics already has TGV packaging capabilities and is expected to apply the products between 2026 - 2027; Jingfang Technology has accumulated many years of mass production experience in the Fan-out process relying on its independent glass substrate technology; Changjiang Electronics Technology and Huatian Technology have both said that they have carried out R & D layouts.

In the niche market, Yicheng Technology was the first to mass-produce the FOMCM platform in 2024, filling the gap in domestic glass panel-level packaging; Xinde Semiconductor and Anjie Technology have made key breakthroughs in 2.5D glass interposers and high-layer (8 + 2 + 8) TGV solutions respectively.

Challenges Faced by Technology and the Industrial Chain

The commercialization of glass substrates cannot be achieved without technological breakthroughs.

Currently, the key path for glass substrate manufacturing is relatively clear. In the core vertical interconnection link, the introduction of laser-induced deep etching (LIDE) technology is a milestone. This process effectively avoids the micro-crack problem of traditional mechanical drilling through a combination of modification and wet etching. Thanks to this, the industry has been able to prepare TGVs with high aspect ratios, laying a physical foundation for improving interconnection density. At the same time, taking advantage of the extremely high nanometer-level flatness of the glass surface, the latest lithography process has been able to achieve redistribution layers with a line width/line spacing of less than 2μm, significantly broadening the data transmission bandwidth between chips.

In addition, through precise control of the glass formula, the thermal expansion coefficient has been successfully locked at 3 - 5ppm/℃. In the large-size packaging experiment of 510mm×515mm, the warpage of the glass substrate has been reduced by more than 50% compared with organic substrates, solving the reliability problem of ultra-large-size chip integration.

However, despite the excellent physical parameters, there are still several real obstacles to overcome in order to transform the technology into a mass production yield with economic benefits. The first is the contradiction between efficiency and filling quality. The throughput of the current mainstream laser drilling technology has not fully met the needs of large-scale production, and during the metallization process of high-aspect-ratio (>15:1) through-holes, copper filling is prone to small voids, which may cause electromigration risks under high current density.

The second is the problem of bonding reliability. Although the thermal matching between glass and chips is good, the difference in thermal expansion coefficients between glass and metal interconnection layers may still cause interface stress concentration during high-temperature processes such as reflow soldering, leading to solder joint failure. In addition, the natural brittleness of glass makes it extremely easy to break during large-panel processing and high-speed transportation, which places extremely high requirements on the automated handling and fixture design of the production line.

Beyond the technical difficulties, the industrial chain also needs to overcome problems such as high supply chain concentration and the lack of industry standards.

The high-purity electronic-grade glass market shows extremely high oligopoly characteristics. Three giants, Corning, Schott, and AGC, firmly control more than 90% of the Low-CTE glass formula and melting technology globally. Considering that the construction period of a glass melting furnace is as long as 12 to 18 months, if the downstream demand explodes in 2026, there will easily be a structural gap in upstream production capacity. This high degree of monopoly also means that packaging factories will be in a weak position in raw material procurement in the initial stage, and the lack of bargaining power will test the cost control capabilities of each company.

In addition, the fragility of glass is the biggest hidden challenge on the mass production line. In the large-size processing and high-speed transportation links, the existing organic substrate production lines cannot be directly reused, and packaging factories must invest heavily to rebuild the handling and fixture systems. What's more troublesome is the lack of industry standards. The non-uniformity of panel sizes, the inconsistent specifications of TGV apertures, and the lag in EDA tool adaptation have jointly created a fragmented ecosystem, driving up the cooperation costs and verification thresholds of the industrial chain.

Conclusion

In summary, 2026 is a critical year for the semiconductor glass substrate industry to transition from technology development to large-scale mass production. Driven by the demand for AI computing power, enterprises in South Korea, Japan, the United States, and the Chinese mainland have all increased their investment, and the global industrial chain is operating at an unprecedented speed.

Although there are still challenges in raw material supply, equipment matching, and yield control, with the maturity of technology and the release of production capacity, glass substrates are expected to gradually establish their position in the high-end semiconductor packaging field. In the foreseeable future, glass may become the cornerstone of computing power infrastructure in the post-Moore era.

This article is from the WeChat public account “Semiconductor Industry Insights” (ID: ICViews), author: Junxi. Republished by 36Kr with authorization.