AI Server Upgrades Drive Triple-Digit Compound Growth in PCB and CCL Market Size in 2026–2028E

Overview

Based on the latest forecasts for AI server rack numbers, PCB and CCL value per system, supplier capacity expansion, and customer allocation, the report raises the global AI server PCB and CCL market size forecasts. The 2027E market sizes are expected to reach US$37.5bn and US$22.1bn, respectively, and further increase to US$83.6bn and US$47.5bn in 2028E. Growth comes from higher server and rack shipments, increased PCB usage, and average selling price increases driven by penetration of high-specification products such as M9, seven-stage or eight-stage HDI, and PCB with more than 30 layers.

Core views

First, GPU and ASIC server demand is growing simultaneously, with ASIC accounting for about 65% of PCB market size in 2026E, while GPU platforms regain about half of the market in 2027–2028E as new architectures ramp up. Second, high-end material upgrades are the core driver of value growth, with six-stage and above HDI, PCB with more than 30 layers, and M9 and above CCL expected to account for 66%, 47%, and 58%, respectively, in 2028E. Third, although nominal capacity is expanding, high-layer-count products occupy more effective capacity and high-end products have lower yields, so capacity utilization is expected to remain high over the next two years. Fourth, the use of M8, M9, or PTFE in backplanes has a relatively limited impact on the overall GPU PCB and CCL market size, but will affect material unit prices and supplier profit allocation. Fifth, suppliers with high-end technology, customer qualifications, automated production lines, and timely capacity expansion capabilities will benefit more.

Analysis framework

The report adopts a bottom-up market size calculation: it first forecasts shipments of general servers and AI servers from leading server brands, ODMs, and other manufacturers, then combines PCB and CCL configurations, unit usage, and average selling prices for GPU, ASIC, and different rack architectures, summarizing market size by application, specification, and supplier allocation; it also sets three material scenarios for the NVL144 Rubin Ultra backplane—M9 base case, PTFE bull case, and M8 bear case—and calculates 12-month target prices for key companies using P/E, enterprise value multiple, and cash return valuation frameworks.

Methodology notes

• Market Size CalculationBottom-Up TAM Calculation

  Server shipments multiplied by value per system

  The global server market is split into general servers, AI server racks, and eight-card-equivalent AI servers, then market size is summarized by GPU and ASIC platforms, PCB and CCL usage, specifications, and average selling prices.

• Sensitivity AnalysisBackplane Material Scenario Analysis

  M9 base case, PTFE bull case, and M8 bear case

  Assuming unchanged backplane shipment volume, the analysis compares the impact of different material routes on backplane average selling prices and the GPU server PCB and CCL market size.

• Relative ValuationP/E and Growth Correlation Valuation

  Target P/E multiplied by forward EPS

  The target P/E ratios for Shengyi Technology and Victory Giant Technology are determined based on the correlation between peer P/E ratios and EPS growth, and applied to 2027E EPS.

• Relative ValuationEnterprise Value Multiple Valuation

  Enterprise value multiples linked to profitability

  Panasonic Holdings uses an EV/EBITDA method based on the relationship between historical enterprise value multiples and margins; Mitsui Kinzoku and Nittobo use materials-sector EV/GCI and CROCI/WACC correlation frameworks.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• Shengyi Technology (600183.SS)

  A core beneficiary of high-end CCL for AI servers and included on the key Buy list.

  Strengths

  It is an important CCL supplier for leading global GPU AI server racks and is expected to upgrade to M9 in 2026–2027E; expansion from GPU to ASIC customers, price increases, and specification upgrades are expected to drive 2026E and 2027E net profit growth of 104% and 73% YoY, respectively.

  Weaknesses

  High growth depends on high-end product qualification, customer allocation, and capacity expansion delivery.

  Comparison

  Compared with general CCL manufacturers, the company has stronger advantages in high-end materials, customer base, and product upgrades; AI server business revenue share is expected to exceed 50% over the long term.

  Risks

  AI infrastructure investment below expectations, customer allocation below expectations, and changes in technology routes.

• Victory Giant Technology (300476.SZ)

  A major beneficiary of high-layer-count PCB and high-end HDI for AI servers, rated Buy.

  Strengths

  Benefits from the ramp-up of global AI infrastructure, increased PCB layer counts, PCB substitution for copper cables, ASIC customer expansion, and highly automated production; Thailand A1 has entered mass production, and A2 is planned to enter mass production in the third quarter of 2026.

  Weaknesses

  New capacity in Thailand still needs to go through construction, yield ramp-up, and customer qualification, with high capital expenditure requirements.

  Comparison

  Automated production lines, warehouse management, and overseas capacity layout enhance its relative competitiveness, with planned output value per unit for A2 and A3 each 40% to 100% higher than A1.

  Risks

  AI server shipment ramp-up slower than expected, specification upgrades slower than expected, and intensifying competition.

• Panasonic Holdings (6752.T)

  A beneficiary of MEGTRON high-end CCL and data center components, rated Buy.

  Strengths

  The MEGTRON series and conductive capacitors have outstanding performance, and MEGTRON 9 has begun small-batch customer validation; the company plans to accelerate investment in data-center-related CCL, BBU, and capacitors.

  Weaknesses

  Group-level decision-making has previously led to slower capacity expansion, and the business portfolio is complex, with non-data-center businesses potentially diluting the contribution from AI growth.

  Comparison

  Its high-performance CCL product capabilities are leading, but its capacity expansion speed and decision-making efficiency lag some focused suppliers.

  Risks

  Capacity expansion delays, business reform falling short of expectations, foreign exchange fluctuations, key customer changes, and intensifying competition in high-end CCL.

• Mitsui Kinzoku (5706.T)

  A beneficiary of high-end electrolytic copper foil and ultra-low-roughness copper foil, rated Buy.

  Strengths

  VSP™ has an estimated global share of about 80% in the ultra-high-end HVLP4 and above market, with deep resin-matching data, process experience, and end-customer trust; revenue CAGR is expected to be 56% in FY26/3–FY29/3.

  Weaknesses

  High-end copper foil capacity may become a bottleneck for demand realization.

  Comparison

  It has established technical barriers that are difficult to replicate in surface smoothness, adhesion balance, and high-end customer qualification.

  Risks

  Slowing AI server demand, yen appreciation, declines in zinc and other metal prices, and insufficient capacity expansion.

• Nittobo (3110.T)

  A core supplier of low-dielectric glass cloth, rated Buy.

  Strengths

  It has significant market and technological advantages in NER Glass; quartz cloth is constrained by cost and yield, which is favorable for NER Glass demand, and revenue CAGR is expected to be 60% in FY26/3–FY29/3.

  Weaknesses

  New capacity will be released gradually only after 2027, and third-generation NEZ Glass remains under development.

  Comparison

  Compared with followers, the technology gap in NER Glass is large, and it is addressing potential quartz cloth substitution through NEZ Glass.

  Risks

  Semiconductor demand fluctuations, exchange rate changes, changes in the competitive landscape for specialty glass, and delays in new product launches.

• Wus Printed Circuit, Shennan Circuits, GCE, EMC, Zhen Ding and TUC

  A Buy-rated portfolio covering the GPU and ASIC server PCB or CCL supply chains.

  Strengths

  They have different customer platforms, product specifications, and regional capacity layouts, and can jointly benefit from AI server demand growth and high-end material upgrades.

  Weaknesses

  The companies differ in customer concentration, supply share, product mix, and capacity expansion progress.

  Comparison

  The portfolio covers multiple segments including finished PCB products, high-end interconnects, and CCL, helping diversify risks from a single platform or single material route.

  Risks

  Changes in customer allocation, price competition, delays in technology qualification, capacity expansion falling short of expectations, and slowing AI capital expenditure.

Key data

• 2027E Global AI Server PCB Market SizeUS$37.5bnPrior forecast was approximately US$27bn; this update raises it by 38%.
• 2027E Global AI Server CCL Market SizeUS$22.1bnPrior forecast was approximately US$19bn; this update raises it by 18%.
• 2028E Global AI Server PCB and CCL Market SizeUS$83.6bn / US$47.5bnCorresponds to 2026–2028E market size CAGRs of 148% and 161%.
• PCB Shipment Volume2026E 1.3 million square meters; 2028E 4.5 million square metersCorresponds to a 2026–2028E CAGR of 85%.
• CCL Shipment Volume2026E 42 million sheets; 2028E 131 million sheetsCorresponds to a 2026–2028E CAGR of 77%.
• Blended CCL Average Selling Price2025 US$93/sheet; 2028E US$362/sheetSpecification upgrades and tight high-end capacity drive continued price increases.
• High-End HDI Share2027E 35%; 2028E 66%Refers to six-stage and above HDI as a percentage of total HDI market size.
• M9 and Above CCL Share2027E 41%; 2028E 58%Shows that high-end low-loss materials are rapidly becoming the main source of value.
• NVIDIA AI Server Racks2027E 92,000 racks; 2028E 148,000 racksForecast includes the next-generation Rubin Ultra architecture.
• Eight-Card-Equivalent AI Server Shipments2026E 1.9 million units; 2027E 2.4 million units; 2028E 2.6 million unitsMainly driven by growth in ASIC server demand.
• NVL144 Rubin Ultra Backplane Chipset Shipments2027E 2,000 sets; 2028E 8,000 setsUsed for backplane material scenario calculations.
• Backplane Bull Case2028E GPU PCB/CCL market size US$42.8bn / US$25.9bnAssumes PTFE is used, with backplane PCB and CCL average selling prices 25% and 30% higher than the base case, respectively.
• Backplane Bear Case2028E GPU PCB/CCL market size US$42.2bn / US$25.3bnAssumes M8 is used, with backplane PCB and CCL average selling prices both 40% lower than the base case.

Impact & implications

AI infrastructure investment is shifting the growth logic of the PCB and CCL industry from pure shipment expansion to simultaneous increases in shipments, content, specifications, and prices. Near-term beneficiaries are manufacturers that have entered the core GPU or ASIC server supply chains and have mass-production capabilities for high-layer-count PCB and high-end CCL; medium-term value will further concentrate in M9 materials, ultra-low-roughness copper foil, low-dielectric glass cloth, and high-end HDI segments. Backplane material selection has a limited impact on the industry's total volume, but has a significant impact on the bargaining power and profit structure of high-end material suppliers. Capacity expansion execution, customer qualification, and yield ramp-up will determine whether companies can convert industry demand into actual revenue and earnings.

Risks

• Global AI infrastructure investment or AI server shipments may fall below expectations.
• Changes in GPU or ASIC platform architectures may lead to adverse changes in PCB and CCL usage and supplier allocation.
• There is uncertainty around the qualification pace and final technology routes for M9, PTFE, quartz cloth, or other materials.
• Yield improvement for high-layer-count PCB, high-end HDI, and high-end CCL may be slower than expected.
• Concentrated supplier capacity expansion may lead to looser supply-demand conditions, lower capacity utilization, or intensified price competition.
• Key customer qualification, overseas factory construction, or mass-production ramp-up may be delayed.
• Key materials such as copper foil, glass cloth, and resin may experience supply bottlenecks.
• Foreign exchange rates, metal prices, and macro demand volatility may affect earnings of Japanese and multinational suppliers.

What to watch

• Final material specifications and mass-production timing for NVIDIA Rubin Ultra and NVL144 backplanes.
• Whether 2027–2028E NVIDIA AI server racks and ASIC server shipments can reach forecasts.
• Penetration speed of M9 and above CCL, six-stage and above HDI, and PCB with more than 30 layers.
• Price increases, capacity utilization, and yield changes for high-end PCB and CCL.
• Shengyi Technology's expansion from GPU customers to ASIC customers and M9 qualification progress.
• Victory Giant Technology's Thailand A2 production start in the third quarter of 2026 and the customer qualification progress of A1 to A3.
• Full-scale mass production of Panasonic MEGTRON 9 in FY3/28 and subsequent additional capacity expansion decisions.
• Whether Mitsui Kinzoku announces large-scale capacity expansion plans for HVLP4 and above copper foil.
• Nittobo's new NER Glass capacity ramp-up and NEZ Glass commercialization progress after 2028.
• Changes in supplier allocation by major cloud service providers and chip platforms.