EN / 中文

Temporary Glass Carriers vs TGV Glass Substrates: Key Differences for AI Chip Packaging

by bandaotichanyeyanjiu·November 13, 2026

I. Clarifying Concepts: What is a Temporary Glass Carrier? Why Has It Been Maturely Applied in Wafer-Level Packaging for Years?

Whenever the industry discusses "glass materials entering AI advanced packaging," capital markets and non-specialist readers often fall into a fundamental cognitive error—confusing the "TGV Permanent Glass Core Substrate," which is on the eve of commercial introduction around 2027–2028, with the "Temporary Glass Carrier" (also known as glass carrier wafer/panel), which has been mass-produced at scale by major domestic and international wafer fabs and OSATs for over a decade. Previously, six listed Chinese companies related to glass substrates issued simultaneous announcements clarifying that TGV glass packaging substrates are still in the early stages of industrialization; this referred specifically to the "glass substrate" requiring millions of micron-level through-glass vias (TGV) filled with copper, ultimately remaining permanently inside the chip. However, the temporary glass carrier, serving as a key consumable and support base in advanced packaging processes, has actually been maturely applied in international and domestic advanced packaging production lines for many years.

So, what exactly is a temporary glass carrier? As illustrated in the engineering cross-section (see Figure 2 below), in processes such as 2.5D silicon interposers, HBM high-bandwidth memory die stacking, and RDL-first/Chip-first Fan-Out Wafer Level Packaging (FOWLP)/Fan-Out Panel Level Packaging (FOPLP), whether thinning12-inch silicon wafers via backside grinding to 25–50μm (thin as cicada wings, only one-third the thickness of A4 paper, prone to curling and breaking like potato chips without support) or directly fabricating sub-micron multi-layer copper Redistribution Layers (RDL) and injecting Epoxy Molding Compound (EMC) without a silicon substrate, a "temporary construction base" with high rigidity and ultra-high flatness is required to support the thin wafer or RDL layer. This sandwich structure, composed of a "bottom high-rigidity glass support carrier (Glass Carrier) + middle photo-thermal/UV release layer (Release Layer) and temporary bonding material (TBM) + top thin wafer/RDL reconstitution layer," constitutes the Temporary Bonding and Debonding (TBDB) system. Once wafer thinning, TSV backside exposure, sub-micron RDL lithography/electroplating, micro-bump placement, and chip mounting are complete, the production line decomposes the release layer via laser irradiation or thermal mechanical methods, fully debonds the glass carrier, cleans it, and recycles it.

II. Why Must Glass Be Used? Three Decisive Physical Advantages of Temporary Glass Carriers Over Silicon Carriers

In the early days of temporary bonding processes, some production lines attempted to use scrapped or downgraded silicon wafers (Silicon Carrier) as temporary carriers. However, as AI computing chip packaging areas expanded exponentially and RDL line widths/space shrank to 2μm or even sub-micron levels, specialized alkali-free borosilicate/aluminosilicate glass carriers promoted by giants like Schott and Corning rapidly completed a comprehensive replacement of silicon carriers, becoming the absolute mainstream in global 2.5D/3D and Fan-Out production lines. The irreplaceability of glass carriers stems from three decisive physical and chemical advantages:

● First, high transmittance from UV to IR (>90%), enabling "laser-transmission second-level lossless debonding" and precise dual-side lithography alignment: This is the core killer feature allowing glass carriers to defeat silicon carriers. Single-crystal silicon is completely opaque to UV light (e.g., 355nm/308nm) and visible light. If silicon carriers were used, debonding would require high-temperature thermal sliding (Thermal Slide, temperatures above 200°C easily scratching large thin wafers) or slow solvent immersion from edge to center (taking hours and unusable for large chips). In contrast, specialized glass carriers have extremely high transmissivity from 300nm UV to 1064nm IR. Debonding equipment can emit high-energy laser beams directly from the back of the transparent glass carrier, precisely hitting the Light-to-Heat Conversion (LTHC) coating (<1μm thick, e.g., 3M™ LTHC solution). The LTHC layer absorbs light and vaporizes/decomposes in nanoseconds, allowing the entire ultra-thin wafer or reconstitution layer to be gently peeled off within seconds with "zero mechanical stress, zero fracture risk." Additionally, the transparency of glass allows lithography machines to identify backside Alignment Marks through the carrier, achieving micron-level dual-side lithography alignment.

● Second, Coefficient of Thermal Expansion (CTE) customizable between 3.2–12.4 ppm/°C, precisely overcoming thermal warpage in heterogeneous integration: Single-crystal silicon CTE is fixed at approximately 2.6 ppm/°C, which matches well when thinning pure silicon TSV wafers. However, in FOWLP/FOPLP or local silicon bridge packaging (e.g., CoWoS-L, SmartPoser-BD), the reconstitution layer contains significant amounts of high-expansion EMC (CTE ~8–15 ppm/°C) and multi-layer copper RDL (copper CTE ~16.5 ppm/°C). Using a fixed-CTE silicon carrier results in severe CTE mismatch between the silicon carrier and the molding layer after high-temperature curing (>200°C) and reflow soldering, causing wafer warpage exceeding several millimeters and direct scrapping. Conversely, manufacturers like Corning can adjust oxide melt formulations to provide glass carriers with continuously tunable CTEs from 3.2 ppm/°C to 12.4 ppm/°C—selecting CTE≈3.2 ppm/°C glass for pure silicon 2.5D interposers to match silicon wafers; selecting high-expansion glass with CTE≈7.0–9.5 ppm/°C for molded reconstituted RDLs to match EMC, physically locking in zero warpage.

● Third, innate "ultra-large square panel" forming genes, sub-micron surface flatness (TTV <1–2μm), and economic viability of recycling: Single-crystal silicon is limited by the physical constraints of Czochralski pulling furnaces, with the global mainstream maximum diameter capped at 12 inches (300mm circular); 500mm or 600mm square single-crystal silicon wafers do not exist. Specialized glass is naturally manufactured using Fusion Draw or float processes, offering extremely low native surface roughness (Ra <0.5 nm), Total Thickness Variation (TTV) controlled within 1–2μm, and easy cutting into 300mm circular wafers or 510×515mm, 600×600mm ultra-large square panels. It also resists strong acid and alkali corrosion, allowing reuse 10–20+ times after cleaning following debonding, with amortized cost per use far lower than silicon carriers.

III. The "Round Wafer Edge Waste Disaster" Driving Shift from 3.3x to 14x Reticle Sizes: Forcing Temporary Glass Carriers "From Round to Square"

If 12-inch (300mm) circular temporary glass carriers are already so mature in wafer-level packaging, why is the global advanced packaging industry currently launching a vigorous "Temporary Glass Carrier From Round Wafer to Square Panel" technology revolution? The answer lies directly in the aggressive expansion of "Reticle Size" by top domestic and international computing power packaging leaders in 2026:

1. SJ Semiconductor (SJ Semi) crosses 3.3x and 6x reticle thresholds: As the largest mainland China supplier of 2.5D silicon interposers, SJ Semi officially released and mass-produced the SmartPoser®-Si 3D silicon interposer platform with3.3x reticle size (~2,830 mm²) on August 31, 2026 (company website disclosed breakthroughs in 0.8μm sub-micron line width and large-size interposer warpage control); shortly after, on October 12, 2026, SJ Semi released the SmartPoser®-BD 2.5D local silicon bridge (Bridge Die) interconnect technology with single-package dimensions reaching 6x reticle (~5,148 mm²) (Eastmoney reports confirmed phased verification completion), supporting ultra-large multi-chiplet packages exceeding 80mm×90mm.

2. Hubei StarTech focuses on 4x reticle 2.5D/3D heterogeneous integration: Incubated by Hubei Jiangcheng Laboratory, with total investment in phases I and II exceeding CNY 7 billion and recently completing an A-round financing of over CNY 4–5 billion, Hubei StarTech Technology Co., Ltd. officially commissioned the first domestic 3D advanced packaging mass production line in Wuhan Optics Valley on September 30, 2026 (reported by Hubei Daily and Wuhan Science and Technology Bureau). Its core focus is a high-performance 2.5D/3D heterogeneous integration platform combining 4x reticle (4x Reticle, ~3,432 mm²) with HBM and 3D TSV.

3. TSMC sprints toward 5.5x and 14x reticle limits: In August 2026, TSMC officially announced mass production of the CoWoS-L advanced packaging solution with 5.5x reticle size (yield breaking 98%–99%, accommodating 12 HBM4 stacks, area approaching 4,700 mm²). Their planned super-packages for 2028–2029 targeting 9x and even 14x reticles push interposer area to a staggering 12,000 mm², with single-package spans exceeding 100mm×120mm.

4. Calculating the "Geometric Dead End" on 300mm Circular Wafers: SJ Semi's 3.3x silicon interposer and 6x reticle silicon bridge, Hubei StarTech's 4x reticle, and TSMC's current CoWoS-L are all built on wafer-level production lines supported by12-inch (300mm) circular temporary glass carriers. However, arranging and cutting huge rectangular packages on circular wafers leads to exponentially worsening Edge Waste when reticle sizes expand beyond 4–6x—the curved edges of circular wafers cannot form complete large rectangles. A 300mm circular wafer with a total area of ~70,685 mm² can yield about 16 modules of 3.3x reticle or 12 modules of 4x reticle; but at 6x reticle (>5,100 mm²), the entire300mm wafer yields only5–8 units (edge waste ratio reaches 45%–55%); once evolving to 14x reticle (12,000 mm²), the entire 300mm wafer yields only 1–2 effective packages, with over 65% of expensive RDL and wafer area turning into edge waste!

Because ultra-large reticle 2.5D/3D heterogeneous integration hit both physical and economic walls on circular wafers, it strongly forces global advanced packaging (such as TSMC's CoPoS panel-level packaging trial production in 2026, and Chengdu ECHINT Technology, the first mainland company to achieve mass production of high-density FO-MCM and recently completing a C-round capital increase of billions) to leap toward 310×310 mm, 510×515 mm, and even 600×600 mm square panel-level packaging (FOPLP/CoPoS)—and the primary foundation supporting this "round-to-square" transition is the large-size square temporary glass carrier!

IV. From 300mm Rounds to 600mm Squares: Four Major Engineering Pain Points in Bonding, Warpage, Laser Debonding, and Reuse

Scaling temporary glass carriers from 300mm diameter rounds (area ~0.07 m²) to 510×515mm (0.26 m²) or even 600×600mm (0.36 m²) square panels is not simple size cropping, but an extreme challenge across four verification dimensions shown in Figure 2: "bonding, warpage control, debonding, contamination control, and repeatability":

1. Bonding Pain Point: How to eliminate the "Edge Bead Effect" during coating of 600mm large square panels and achieve TTV <2μm? On 300mm circular wafers, Temporary Bonding Material (TBM) and LTHC release layers can be evenly spread via centrifugal force using high-speed Spin Coating. However, on 600×600mm square glass carriers, high-speed spin coating causes severe adhesive thickness accumulation and bubble defects at the four right-angle corners due to air turbulence, pushing Total Thickness Variation (TTV) beyond 10μm, subsequently causing defocus in subsequent step-and-repeat lithography for 0.8–2μm line width RDLs. Engineering Solution: Panel-level lines must abandon pure spin coating in favor of high-precision Slot-Die Coating or Dry Film Vacuum Lamination, combined with multi-zone flexible airbag array vacuum hot-press bonding machines, strictly compressing the post-bonding TTV of the entire 600mm glass carrier to ≤1.5–2.0μm.

2. Warpage Pain Point: How to tame the "Bi-Axial Asymmetric Saddle-Shaped Warpage" of up to 10–15mm in square panels? Stress distribution in circular wafers is centrally symmetric; however, square large panels experience intense bi-axial shear stress accumulating at the four right angles during multi-layer high-density copper RDL electroplating and large-area EMC high-temperature curing shrinkage. Free-state warpage of 600mm panels can reach 10–15mm, causing automated handling robots to jam or glass carriers to crack. Engineering Solution: On one hand, adopt high Young's modulus (>70–80 GPa) custom-CTE glass carriers thickened from wafer-level 0.7mm to 1.1mm–2.0mm (e.g., precisely tuning glass CTE to 7.0–8.8 ppm/K to lock in molding layer shrinkage rate); on the other hand, introduce compensating stress films on the back of glass carriers or use double-sided clamped annealing processes, controlling dynamic full-process warpage of large panels within the safe limit for lithography and mounting of ≤0.5–1.0mm.

3. Laser Debonding Pain Point: How to achieve millisecond-level uniform laser peeling over a 0.36 square meter large format without damaging fragile micro-bumps? On large panels carrying dozens of 4x–6x reticle AI chiplets, any mechanical prying or thermal sliding triggers catastrophic local fragmentation; UV (355nm/308nm) or IR (1064nm) laser-transmission debonding is the only viable mass production route. However, on a 600×600mm giant format, if the energy in laser spot overlap regions is too high, it penetrates the LTHC layer and burns the top RDL pads; if too low, it leaves un-vaporized "Undebonded Spots," tearing micrometer-thick RDL copper wires the instant the glass carrier is lifted. Engineering Solution: Employ Flat-top Beam Shaping + High-Speed Galvo Flying Scan System, combined with high-sensitivity low-threshold LTHC photosensitive release coatings, to complete dead-spot-free uniform vaporization peeling of the entire panel within 30–60 seconds.

4. Contamination & Reusability Pain Point: How to achieve zero carbon residue and >15 cycles of cost reduction for glass carriers? After laser vaporization of the LTHC release layer, micron-level Carbon Soot and polymer residues often remain on the thin wafer/RDL surface and the glass carrier. Direct entry into electroplating or mounting leads to open circuits in Micro-bumps; given the high cost of precision large-size specialized glass carriers, inability to recycle them erodes panel-level packaging profits. Engineering Solution: Develop accompanying eco-friendly specialized wet stripping cleaning solutions (e.g., Feikai Materials, 3M matching cleaners) combined with Plasma Ashing processes, achieving "zero residue, zero metal ion contamination" on the wafer side while restoring the surface roughness of the debonded glass carrier losslessly, realizing high-yield cyclic reuse of a single glass carrier for 10–20+ times.

V. Current Application Status and Full Industry Chain Breakthrough Map of Domestic and International Temporary Glass Carriers (Materials-Equipment-Lines)

Looking at the global and mainland China advanced packaging ecosystem in 2026, surrounding the industrial wave of "wafer-level mature volume ramp-up + panel-level accelerated introduction" for temporary glass carriers, domestic and international players have formed clear competitive and cooperative landscapes across four segments: glass carrier substrates, temporary bonding/release materials, TBDB bonding/laser debonding equipment, and downstream packaging/testing lines:

1. Glass Carrier Substrate Segment: Internationally dominated long-term by US Corning (providing full series of semiconductor glass wafers and panels with CTE 3.2–12.4 ppm/°C adjustable), Germany Schott (established a semiconductor advanced packaging glass solutions department in Suzhou, providing Borofloat® 33 and D263® series high-transparency low-TTV carriers), Japan AGC, and NEG (Nippon Electric Glass). Domestically, local manufacturers such as WG Tech (subsidiary TGVTECH), Triumph Science & Technology, Rainbow Shares, GobiGa, and Qibin Group are leveraging their expertise in ultra-thin optoelectronic glass fusion draw and precision cold processing to accelerate penetration from display panel glass to high-flatness semiconductor temporary glass carriers (300mm rounds and 510×515mm/600×600mm squares).

2. Temporary Bonding Material (TBM) & Light-to-Heat Release Layer (LTHC) Segment: Internationally held by US 3M (classic LC3200 UV adhesive and LTHC coating systems), Brewer Science, Japan Shin-Etsu Chemical, and Tokyo Ohka Kogyo (TOK), occupying over 80% share. Domestic material leaders have achieved substantive mass production breakthroughs—Feikai Materials (300398.SZ) has developed a complete solution including temporary bonding adhesives, LTHC photosensitive adhesives, and matching cleaning fluids, fully supporting thermal, mechanical, and laser debonding; Hubei Dinglong (300054.SZ)'s self-developed DT series temporary bonding adhesives have secured procurement orders from mainstream domestic wafer fabs and entered batch delivery (Sina Finance confirms breaking overseas import monopoly); additionally, Yunzhuo New Materials, Bayi Shikong, and Strong New Materials are accelerating introduction in HBM and 2.5D/FOPLP temporary bonding materials.

3. Temporary Bonding & Laser Debonding Equipment (TBDB) Segment: Internationally controlled by Austrian EV Group (EVG) and German SUSS MicroTec duopoly. In the domestic equipment breakthrough battle, Kingsemi (688037.SH), leveraging its expertise in coating/developing and wafer alignment, has secured mass production and verification orders for its self-developed temporary bonders and laser/mechanical debonders from multiple customers (Tonghuashun discloses significant competitive advantage in temporary bonding equipment); Piotech (688072.SH) deeply cultivates wafer bonding and chip reconstitution bonding equipment series; simultaneously, iSABers, MaxWell, Xinrui Technology, and Han's Laser/HGLASER are launching a full assault from 12-inch wafer-level TBDB equipment to 510×515mm/600×600mm panel-level large-format laser debonding stations.

References:

[1] SJ Semi Official Website: "SJ Semi's 6x Reticle Size 2.5D Silicon Bridge (SmartPoser®-BD) Completes Verification and Debuts at 2026 China Industrial Expo Guoxin Exhibition" (2026-10-20): https://www.sjsemi.com/about/22699/155393.html

[2] SJ Semi Official Website: "Expanding Chip Integration Dimension Boundaries: SJ Semi Releases 3.3x Reticle Size Silicon-Based 2.5D Packaging Platform SmartPoser®-Si" (2026-08-31): https://www.sjsemi.com/about/22699/155287.html

[3] Eastmoney: "SJ Semi: 2.5D Silicon Bridge Technology with Single Package Size Reaching 6x Reticle Completes Phased Verification" (2026-10-13): https://finance.eastmoney.com/a/202610133894181211.html

[4] Hubei Daily: "Another Step Towards Autonomous Control of Advanced Packaging for High-End Chips in China: First Domestic 3D Advanced Packaging Mass Production Line Commissioned at Hubei StarTech" (2026-09-30): https://news.hubeidaily.net/mobile/c_6033359.html

[5] CLS: "Over CNY 4 Billion in A-Round Financing, Wuhan Optics Valley's Hubei StarTech Emerges as 'Dark Horse' in Advanced Packaging" (2026-07-07): https://www.cls.cn/detail/2419500

[6] TradingKey / TSMC OCP Summit: "TSMC's 5.5x Reticle CoWoS Yield Breaks 98% Entering Mass Production, Targeting 14x Reticle Size by 2029" (2026-08-12): https://www.tradingkey.com/zh-hans/analysis/stocks/us-stock/262101030-tsmc-cowos-yield-breakthrough-14x-reticle-roadmap-tradingkey

[7] CommonWealth Magazine: "What is CoPoS? TSMC Mass Produces 5.5x Reticle CoWoS, Actively Developing 9x and 14x Reticle and Panel-Level Packaging" (2026-10-21): https://www.cw.com.tw/article/5140709

[8] National Business Daily: "Chengdu ECHINT Technology, First Mainland Enterprise to Achieve Mass Production of High-Density Panel-Level FO-MCM, Completes Multi-Billion Yuan C-Round Capital Increase" (2026-11-05): https://www.nbd.com.cn/articles/2026-11-05/4639868.html

[9] Corning Official Technical Library: "Semiconductor Glass Wafers (CTE 3.2~12.4 ppm/°C Adjustable)" : https://www.corning.com/cn/zh/products/advanced-optics/product-materials/semiconductor-laser-optic-components/semiconductor-glass-wafers.html

[10] SCHOTT Group Official White Paper: "High-Performance Glass Carriers: Temporary Bonding and Laser Debonding Carriers for Silicon Wafers and Chips": https://media.schott.com/api/public/content/2d2482b398304345a734b11b51b4b393?v=2075a78e&download=true

[11] 3M China Semiconductor Solutions Center: "3M™ Temporary Bonding and Laser Debonding Solutions (LC3200 UV Adhesive and LTHC Coating)": https://www.3m.com.cn/3M/zh_CN/semiconductor-cn/solutions/advancedpackaging/tbdb/

[12] Feikai Materials Official Website: "Wafer-Level Packaging Materials: Complete Temporary Bonding and Laser Debonding Solution Including Bonding Adhesives, Photosensitive Adhesives, and Cleaning Fluids": https://www.phichem.com.cn/pro2-2/462.html

[13] Hubei Dinglong Official Website and Announcements: "DT Series Temporary Bonding Adhesive Product Details and Announcement of First Procurement Order from Mainstream Domestic Wafer Fab Customers": https://www.dl-kg.com/product_detail/34.html

[14] Phoenix Finance / Tonghuashun: "Kingsemi: Good Performance of New Mass Production and Verification Orders for Temporary Bonders and Debonders Applied in Chiplet Fields": https://finance.ifeng.com/c/8jhNeBPpMk5

[15] CLS: "Six Listed Companies Issue Simultaneous Announcements Clarifying That TGV Glass Substrate Related Technologies Are Still in Early Industrialization Stages" (2026-06-18): https://www.cls.cn/detail/2403170