Report Date: September 28, 2026. Analyst: Wu Qidi, Registration No. A0190523020001. Analyst: Song Jinzhi, Registration No. A0190526020002.

Key Investment Points

The global CCL market exceeds RMB 100 billion, with high-end segment share steadily rising. Copper Clad Laminate (CCL) is manufactured by impregnating glass fiber cloth with resin, covering one or both sides with copper foil, and hot-pressing the assembly. It serves three core functions in PCBs: conduction, insulation, and structural support, making it the segment with the highest technical barriers and greatest profit elasticity in the PCB supply chain. On the cost side, direct materials account for approximately 60% of PCB costs, with CCL representing about 30% and serving as the core cost center. CCL's own cost structure comprises copper foil (42%), resin (26%), and glass fiber cloth (19%). According to Prismark, the global CCL market was approximately RMB 100 billion in 2024, with the top four manufacturers—Kingboard, Shengyi Technology, TUC, and Nanya New Material—holding a combined share of about 50%. End products are evolving toward thinner, lighter, smaller form factors with high-frequency and high-speed capabilities, forcing CCL to upgrade toward high-frequency high-speed tracks: revenue from M7 and above specialty CCLs reached USD 5.705 billion in 2024, up 40% year-over-year, with its share jumping from 32% in 2023 to 38%. M7-M9 grade products are primarily used in AI server high-speed backplanes and 800G/1.6T optical modules, with per-unit value reaching several times that of standard FR-4. High-endization has become the core growth engine of the industry.

AI Server Iteration Drives Both Volume and Price Increases for CCL

AI server acceleration drives simultaneous volume and price increases for CCL. On the pricing side, NVIDIA's CCL platform has upgraded from GB200's M4+M8 to Rubin's M8/M9: Switch boards reach 40+ layers at M9 grade paired with Q cloth, Midplane boards reach 44 layers at M9 grade, Rubin Ultra orthogonal backplanes reach 78 layers, and the next-generation Feynman backplane is expected to reach 144 layers, potentially doubling layer counts; Google TPU v8 and Amazon Trainium3 are also migrating from M7 to M8. On the volume side, increasingly complex rack architectures drive per-unit consumption higher: Hopper racks use approximately 212 sheets, GB200/GB300 use about 164 sheets, and VR200 uses about 578 sheets, representing roughly 1.7x and 2.5x increases compared to Hopper and GB200 respectively; on the ASIC side, Google TPU per-server usage has risen from 144 sheets in V6 to 200 sheets in V8. We estimate CCL demand for 2025-2027 at 22.56 million, 56.65 million, and 118.99 million sheets respectively, representing year-over-year growth of +4%, +151%, and +110%. The volume-price double click drives sustained industry prosperity.

Core Raw Materials for CCL

High-end CCL volume ramp transmits bottlenecks upstream, with electronic cloth and copper foil forming a dual constraint. Electronic cloth upgrades progressively from E cloth to NE cloth, L cloth, and Q cloth, with prices rising accordingly. High-end electronic cloth can essentially only be produced on Toyota JAT910 weaving machines, with Toyota's annual capacity at 2,400 units and orders booked through 2030; Low-Dk cloth efficiency is only 40% that of FR-4. Electronic yarn net new capacity for 2026-2028 is only 116,000/36,000/78,000 tons, with no new capacity ignition in 2027, and G75 electronic yarn rising from RMB 9,000/ton to RMB 16,850/ton. On the copper foil side, M6 and above require HVLP copper foil, with the Rubin platform upgrading to HVLP4 (roughness 0.4-0.6μm), domestically quoted at approximately RMB 180,000/ton, far exceeding standard foil at RMB 17,000-19,000. High-end HVLP is dominated by Mitsui Mining (approximately 50% share), with HVLP4 available from only three suppliers, and 3μm ultra-thin copper foil also monopolized by the same. Tongguan Copper Foil has achieved HVLP mass shipments first, while Defu and others are in the verification stage. Domestic substitution is accelerating, laying the prerequisite for domestic CCL breakthroughs at M8 and above.

Related Targets

1) PCB manufacturers: Victory Giant Technology, Dongshan Precision, Pengding Holdings, Jingwang Electronics, WUS Printed Circuit, Guanghe Technology, Shengyi Technology, Shiyun Circuit. 2) CCL and electronic cloth: Shengyi Technology, Shengyi Electronics, Nanya New Material, China Jushi, Sinoma Science & Technology, Honghe Technology, International Composite Materials.

Risk Warnings

AI server demand falling short of expectations; raw material price volatility risk; high-end product technology breakthroughs and capacity release falling short of expectations; sustainability of price increases.

PCB Index and CSI 300 Index Trend. Source: Wind, Yuanda Information Securities Research Institute.

Report Body: I. CCL Market Size Exceeds RMB 100 Billion, High-End Share Continues to Rise

1.1 CCL is the Most Core Substrate for PCBs

Copper Clad Laminate (CCL) is a sheet material made by impregnating glass fiber cloth with resin, covering one or both sides with copper foil, and hot-pressing. It serves three major functions in PCBs: conduction, insulation, and support. According to Prismark, direct materials including CCL, copper foil, and phosphor copper balls account for approximately 60% of PCB costs, with CCL representing about 30%, making it the core cost center of PCBs. CCL's three core raw materials are copper foil, resin, and glass fiber cloth, accounting for 42%, 26%, and 19% of CCL costs respectively. Copper foil prices mainly depend on copper prices, while glass fiber cloth prices are significantly influenced by supply-demand dynamics. CCL directly benefits from the rapid prosperity improvement brought by AI server end-demand expansion: the upstream of the CCL supply chain consists of core raw materials including copper foil, electronic resin, and electronic glass fiber cloth; midstream CCL manufacturing is the segment with the highest technical barriers and greatest profit elasticity; downstream consists of PCB manufacturing and packaging substrates. Figure 3: CCL Supply Chain Upstream-Downstream Breakdown. Source: Company announcements, Prismark, Yuanda Information Securities Research Institute.

CCL industry concentration is relatively high. According to Prismark, the CCL industry size was approximately RMB 100 billion in 2024, with the top four suppliers being Kingboard, Shengyi Technology, TUC, and Nanya New Material, with market shares (by revenue) of 14%, 14%, 13%, and 9% respectively, totaling approximately 50%. Figure 4: 2024 CCL Global Industry Landscape (by revenue). Source: Prismark, Yuanda Information Securities Research Institute.

1.2 CCL Moving Toward High-Endization, M7 and Above Growing Rapidly

End products continue to evolve toward thin, light, small form factors and high-frequency high-speed capabilities, forcing PCBs to iterate toward high density and high performance, which in turn drives CCL upgrades toward high-frequency high-speed tracks. In high-speed application scenarios, with 5G communication technology iteration, signal frequency and transmission rates have risen significantly, with theoretical rates reaching 10-20Gbps, requiring supporting CCLs to have at least mid-to-low-grade dielectric loss characteristics. For CCL materials, the lower the Df (dielectric loss) value, the higher the technical barriers to material R&D and manufacturing. Figure 5: High-Speed Interconnect Technology Places Higher Demands on CCL Materials. Source: "224G High-Speed Interconnect Demand and Challenges for PCB and CCL", Yuanda Information Securities Research Institute.

Benefiting from AI demand expansion, M7 and above specialty CCLs are growing rapidly. The industry uses Panasonic's M series as a universal technical benchmark: higher grades correspond to lower dielectric constant Dk and dielectric loss Df, and can accommodate higher rates. M7-M9 are used in AI server high-speed backplanes and 800G/1.6T optical modules, with per-sheet prices several times higher than standard FR-4. In 2024, specialty CCLs (M7 and above) accounted for 38% share at USD 5.705 billion, growing 40% year-over-year; standard FR-4 was USD 4.354 billion (29% share), growing only 1%. Specialty board share jumped from 32% in 2023 to 38% in 2024. Table 1: Different CCL Grades with Corresponding Applications and Major Manufacturers. Source: wccftech, Yuanda Information Securities Research Institute.

II. AI Server Iteration Drives Both Volume and Price Increases for CCL

2.1 Price: Grade Escalation Generation by Generation

AI server iteration drives CCL value growth: NVIDIA's CCL platform upgrades from GB200's M4+M8 to Rubin's M8/M9: Switch board at 40+ layers, M9 grade with Q cloth; Midplane board at 44 layers (2×22), M9 grade; Rubin Ultra orthogonal backplane at 78 layers (3×26), using TUC's EM-896K3 (M9); next-generation Feynman orthogonal backplane expected to reach 144 layers (6×24), with layer counts potentially doubling. Google TPU v8 and Amazon Trainium3 are simultaneously migrating from M7 to M8. Table 2: CCL Configuration Evolution for NVIDIA and Cloud Providers' Core Boards. Source: NVIDIA official website, JPM, Yuanda Information Securities Research Institute.

2.2 Volume: Architecture Changes Bring New Incremental Usage

Architecture changes bring new incremental usage: Hopper single rack uses approximately 212 CCL sheets (Compute 16 motherboards + Switch 5 motherboards × 12 sheets/board/4); GB200/GB300 approximately 164 sheets; VR200 approximately 578 sheets (Compute 36 motherboards × 5 sheets/board + Switch 9 motherboards × 40 sheets/board + backplane 38 sheets), representing approximately 1.7x increase over Hopper and approximately 2.5x increase over GB200. On the ASIC side, taking Google TPU as an example, per-server usage has risen from 144 sheets in V6 to 200 sheets in V8. We estimate: CCL demand for 2025-2027 at 22.56 million, 56.65 million, and 118.99 million sheets respectively, growing +4%, +151%, and +110% year-over-year. The core growth drivers come from AI server shipment expansion and incremental substrate demand generated by increasingly complex next-generation computing hardware architectures. Table 3: AI Server Demand-Driven CCL Demand Calculation. Source: NVIDIA official website, JPM, Yuanda Information Securities Research Institute.

III. CCL Core Raw Materials

3.1 Electronic Cloth: Dual Constraints from Weaving Machines and Electronic Yarn

Electronic cloth is classified by performance into: standard E cloth, Low-Dk first-generation cloth (NE cloth), Low-Dk second-generation cloth (L cloth), Low-CTE cloth, and Q cloth (quartz cloth). Dk decreases from 6.2-6.6 for E cloth to 3.5-3.8 for Q cloth, Df decreases from approximately 0.006 to approximately 0.0002, while prices rise from approximately RMB 8/meter for standard E cloth to approximately RMB 26/meter for first-generation cloth, RMB 150-160/meter for second-generation cloth, and over RMB 200/meter for Q cloth. According to Kingboard Laminates announcements, the company has raised prices 9 times since August 2025, with cumulative FR-4 price increases approaching 70% in 2026. Table 4: Electronic Cloth Types, Applications, and Major Manufacturers. Source: Kingboard official website, Yuanda Information Securities Research Institute.

Electronic cloth supply constraints come from two segments: 1) Weaving machine equipment monopoly plus efficiency losses. High-end electronic cloth can essentially only be produced on Toyota JAT910 weaving machines, with Toyota's annual capacity at 2,400 units with no expansion plans, orders booked through 2030, and delivery lead times of 18-24 months. In terms of industry concentration, Toyota holds 50%, Tsudakoma 20%, and domestic players (Rifa, Taitan, etc.) 30%, but domestic weaving machines still have generational gaps in core processes such as tension control, with market entry expected only after 2027. More critically, there are efficiency losses: Low-Dk cloth production efficiency is only 40% of FR-4, and switching from E cloth to AI high-end cloth involves approximately 60% efficiency loss. 2) Electronic yarn new capacity is nearly stagnant: net new actual capacity for electronic yarn in 2026-2028 is 116,000/36,000/78,000 tons respectively, corresponding to +10.8%/+3.1%/+6.4% year-over-year for operating capacity; no new capacity ignition in 2027, with some relief only after China Jushi's 50,000-ton capacity ignition in 2028, and some electronic yarn switching to AI production results in even lower standard yarn output growth. On pricing, G75 electronic yarn rose from RMB 9,000/ton in September 2025 to RMB 16,850/ton in July 2026, with month-over-month increases from May to July at times exceeding 7628 electronic cloth. Figure 6: G57 Electronic Yarn and 7628 Electronic Cloth Prices. Source: Zhuochuang Information, Yuanda Information Securities Research Institute.

3.2 Copper Foil: High-End HVLP Supply Concentrated, Domestic Substitution Accelerating

The core driver of copper foil technology upgrade is the skin effect—the higher the signal frequency, the more current concentrates on the conductor surface, demanding stricter copper foil surface roughness requirements. M6 and below use RTF, M6 and above use HVLP; GB200/GB300 use HVLP3, Rubin upgrades to HVLP4 (roughness 0.4-0.6μm). Processing fee premiums across generations are significant: HTE standard foil at RMB 17,000-19,000/ton, RTF at RMB 25,000-37,000/ton, HVLP4 domestically quoted at approximately RMB 180,000/ton. On the supply side: global PCB copper foil total capacity is approximately 1.2 million tons (China accounts for 70%), but high-end HVLP is dominated by Mitsui Mining (approximately 50% global share), with HVLP4 available from only Circuit Foil, Mitsui Mining, and SK Nexilis; 3μm ultra-thin copper foil is monopolized by Mitsui Mining, with domestic mass production still far off. On the domestic front, Tongguan has achieved HVLP mass shipments first (2024 H2), while Defu and Jinbao are in the verification stage—this is a necessary prerequisite for domestic CCL breakthroughs to M8 and above grades. Table 5: Copper Foil Classification, Application Scenarios, and Core Manufacturers. Source: Company announcements, Prismark, Yuanda Information Securities Research Institute.

IV. Related Targets

4.1 Related Targets

1) PCB manufacturers: Victory Giant Technology, Dongshan Precision, Pengding Holdings, Jingwang Electronics, WUS Printed Circuit, Guanghe Technology, Shengyi Technology, Shiyun Circuit. 2) CCL and electronic cloth: Shengyi Technology, Shengyi Electronics, Nanya New Material, China Jushi, Sinoma Science & Technology, Honghe Technology, International Composite Materials.

4.2 Wind Earnings Forecast

Table 6: Key Companies Wind Earnings Forecast. Source: Wind Consensus Estimates (2026/09/24), Yuanda Information Securities Research Institute.

V. Risk Warnings

AI server demand falling short of expectations; raw material price volatility risk; high-end product technology breakthroughs and capacity release falling short of expectations; sustainability of price increases.

Investment Rating Explanation

Analyst Declaration

The author holds a securities investment consulting practice qualification granted by the Securities Association of China and is registered as a securities analyst, producing this report with a diligent professional attitude, independently and objectively. The analytical logic is based on the author's professional understanding, and this report clearly and accurately reflects the author's research views. No part of the author's compensation has been, is, or will be directly or indirectly related to the specific recommendations or views in this report. This is hereby declared.

Important Declaration

Hebei Yuanda Information Technology Co., Ltd. holds a securities investment consulting business qualification, with securities business license number: 911301001043661976. This report is issued only in mainland China and is for the use of clients of Hebei Yuanda Information Technology Co., Ltd. (hereinafter referred to as "the Company") only. The Company will not treat recipients as clients by virtue of their receiving this report. The information in this report is sourced from publicly available materials, and the Company makes no guarantee as to the accuracy and completeness of such information, nor does it guarantee that the information and suggestions contained herein will not change. The Company has endeavored to make the report objective and fair, but the views, conclusions, and suggestions herein are for reference only and do not constitute the author's definitive judgment on securities price movements or market trends. The information or opinions expressed in this report do not constitute investment advice to any person, and investors should independently evaluate the information and opinions in this report. This report reflects only the Company's judgment on the date of publication. At different times, the Company may issue other reports that are inconsistent with or have different conclusions from the information contained in this report; the different views, opinions, and analytical methods reflected in this report do not represent the position of the Company or other affiliated institutions. Meanwhile, the Company may modify the information contained in this report without notice, and investors should pay attention to relevant updates or modifications on their own. The Company and the author, to the best of their knowledge, have no conflicts of interest with the securities evaluated or recommended in this report that require disclosure or restrictive or silent measures under laws and regulations. The copyright of this report belongs solely to the Company. Without written permission, no institution or individual may reproduce, copy, or distribute this report in any form. If quoted, the source must be indicated as Yuanda Information Securities Research Institute, and no quotations, deletions, or modifications contrary to the original meaning of this report are permitted. When publishing or forwarding this securities research report or its abstract, the publisher and publication date of this report shall be indicated, and the risks of using the securities research report shall be noted. For unauthorized publication or forwarding of this report, the Company reserves the right to pursue legal liability.

Financial Statements and Indicator Forecasts