AI Server Volume and Pricing Rise Together; Goldman Sachs Expects Global AI PCB/CCL Markets to Reach US$84bn and US$48bn by 2028E

Overview

The report recalculates market opportunities based on global AI server rack demand, PCB/CCL value per system, industry capacity expansion plans, and supplier shares. Goldman Sachs believes AI infrastructure buildout is not only driving rapid shipment growth but will also promote penetration of high-end specifications such as M9 and above CCL, seven-step or eight-step HDI, and 30-layer and above PCBs, creating a significant volume-price resonance. Even as major manufacturers continue expanding capacity, high-end products consume more capacity and have lower yields, so effective capacity is expected to remain tight over the next two years.

Core views

The core conclusions include: first, global AI PCB/CCL demand growth is significantly higher than previously forecast, with 2027E market sizes raised by 38% and 18%, respectively; second, both GPU and ASIC servers will contribute to growth, accounting for approximately 51% and 49% of the 2028E PCB market, respectively; third, value growth comes not only from increases in the number of servers and racks, but also from higher layer counts, HDI upgrades, M9 material adoption, and more PCB replacement of copper cables; fourth, backplane material scenarios have a relatively limited impact on the overall GPU PCB/CCL market size, but the unit value of the PTFE solution is higher than M9, while the M8 solution is significantly lower; fifth, suppliers with high-end product technologies, customer qualifications, capacity expansion capabilities, or monopoly advantages in key materials will benefit first.

Analysis framework

The report uses a bottom-up market size model, breaking the global server market into general-purpose servers, AI server racks, and eight-card-equivalent AI servers, and combines server shipments, PCB/CCL usage per system, and average selling prices to estimate market size. It also cross-validates by GPU and ASIC platforms, product applications, technical specifications, and supplier allocation, and conducts scenario analysis for NVL144 Rubin Ultra backplanes using M9, PTFE, or M8 materials.

Methodology notes

• Market Size EstimationBottom-Up Server Market Model

  Shipments multiplied by value per unit

  The global server market consists of general-purpose servers, AI server racks, and eight-card-equivalent AI servers; server shipments are aggregated from major brands, ODM direct supply, and other manufacturers, while revenue and PCB/CCL market size are then calculated from shipments and average selling prices.

• Scenario AnalysisBackplane Material Scenario Analysis

  Base, bull, and bear scenarios

  Assuming NVL144 Rubin Ultra uses a backplane and shipment volume remains unchanged, the base scenario uses M9, the bull scenario uses higher-value PTFE, and the bear scenario uses lower-value M8 to measure the impact of changes in material routes on market size.

• Equity ValuationPeer Growth and Valuation Correlation

  Target P/E method

  The target prices for 600183.SS and 300476.SZ use target P/E multiples of 50.4x and 26.3x, respectively, and are determined in conjunction with the companies' expected EPS growth and peer valuation relationships.

• Equity ValuationEnterprise Value and Cash Return Valuation

  EV/EBITDA and EV/GCI-CROCI/WACC frameworks

  6752.T is valued at 9.0x EV/EBITDA; target prices for 5706.T and 3110.T are determined based on the correlation between materials industry EV/GCI and expected CROCI/WACC.

Asset mapping & comparison

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

• Shengyi Technology（600183.SS）

  A direct beneficiary in high-end CCL and included on Goldman Sachs' Conviction Buy List.

  Strengths

  It is an important supplier to leading global GPU AI server racks and is expected to migrate to M9-grade CCL in 2026E-2027E; expansion from GPU to ASIC customers, specification upgrades, and price increases are expected to drive 2026E/2027E net profit growth of 104%/73% YoY, with operating margin expected to rise from 15% in 2025 to 20% in 2027E.

  Weaknesses

  Earnings delivery depends on high-end product qualifications, yield improvement, capacity ramp-up, and customer allocation.

  Comparison

  Compared with general CCL suppliers, the company stands out more in high-end server materials, customer coverage, and earnings leverage.

  Risks

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

• Victory Giant（300476.SZ）

  A direct beneficiary in advanced PCB for AI servers.

  Strengths

  Benefits from global AI infrastructure volume growth, higher-layer-count PCBs, PCB replacement of copper cables, and ASIC customer expansion; automated production and the Thailand A1, A2, and A3 bases provide capacity expansion and geographic diversification capabilities.

  Weaknesses

  New capacity in Thailand still requires construction, mass production, and customer qualification, creating short-term ramp-up uncertainty.

  Comparison

  Compared with peers, the company has advantages in AI PCB customer coverage, automation, and overseas capacity expansion.

  Risks

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

• Panasonic Holdings（6752.T）

  Benefits through the MEGTRON series CCL and data center-related products.

  Strengths

  MEGTRON 9 has begun small-lot customer validation and is planned for full mass production from FY3/28; the share of high-performance products is expected to rise to about 60% in FY3/27, with capacity expansion plans in Guangzhou, Suzhou, and Ayutthaya.

  Weaknesses

  The group has historically been relatively slow in timely capacity expansion decisions, and its conglomerate structure may affect investment execution efficiency.

  Comparison

  Product performance is leading, but the speed of growth delivery depends more on group reform and capex decisions.

  Risks

  Reform progress, exchange rates, global demand, competition, investment burden, and data center-related product capacity expansion falling short of expectations.

• Mitsui Kinzoku（5706.T）

  Supplier of ultra-low-roughness electrolytic copper foil for high-end PCBs.

  Strengths

  The VSP series addresses the trade-off between low transmission loss and adhesion; the company has an estimated global share of about 80% in the HVLP4 and above high-end market and possesses deep material matching data and customer trust; VSP revenue is expected to compound at 56% over FY26/3-FY29/3.

  Weaknesses

  High-end copper foil capacity may become a growth bottleneck, requiring further large-scale capacity expansion.

  Comparison

  Its share, quality, and accumulated co-development capabilities in the HVLP4 and above fields are significantly ahead of competitors.

  Risks

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

• Nitto Boseki（3110.T）

  Supplier of low-dielectric glass cloth required for AI server CCL.

  Strengths

  It has an overwhelmingly leading share and clear technological advantages in the NER Glass market; NER Glass revenue is expected to compound at 60% over FY26/3-FY29/3; it is expanding capacity and developing third-generation NEZ Glass to address higher-performance demand.

  Weaknesses

  Competition in first-generation NE Glass is relatively intense, and new capacity and commercialization of NEZ Glass will take time.

  Comparison

  Compared with competitors such as Asahi Kasei, the company has stronger technological and share advantages in second-generation NER Glass.

  Risks

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

• 002463.SZ、002916.SZ、2368.TW、2383.TW、4958.TW、6274.TWO

  Other PCB or CCL supply-chain Buy-rated stocks listed by Goldman Sachs.

  Strengths

  Expected to benefit from AI server shipment growth, high-end material penetration, higher unit value, and reallocation of supplier shares.

  Weaknesses

  The provided text discloses limited company-specific operating data and valuation basis for each company.

  Comparison

  Each company's benefit path differs across PCB, CCL, customer platforms, and regional capacity layout, and should be tracked in conjunction with the supplier allocation table.

  Risks

  Customer concentration, competition, qualification progress, capacity expansion execution, yield, and changes in technology routes.

Key data

• 2027E Global AI PCB Market SizeUS$37.5bn, approximately US$38bnRaised 38% from the previous forecast of US$27bn.
• 2027E Global AI CCL Market SizeUS$22.1bn, approximately US$22bnRaised 18% from the previous forecast of US$19bn.
• 2028E Global AI PCB/CCL Market SizeUS$83.6bn/US$47.5bnRounded in the report summary to US$84bn/US$48bn.
• 2026E-2028E Market Size Compound GrowthPCB 148%; CCL 161%Driven jointly by specification upgrades, increased content per system, and shipment growth.
• 2026E-2028E PCB ShipmentsIncrease from 1.3 million square meters to 4.5 million square metersCorresponding to an 85% CAGR.
• 2026E-2028E CCL ShipmentsIncrease from 42 million sheets to 131 million sheetsCorresponding to a 77% CAGR; the detailed table shows 131 million sheets as the 2028E figure.
• Blended CCL Average Selling Price2026E US$165/sheet, 2027E US$282/sheet, 2028E US$362/sheetHigh-end material penetration and tight supply-demand conditions support price increases.
• Advanced HDI Penetration RateSix-step and above accounted for 35% in 2027E and 66% in 2028ESpecification upgrades significantly increase PCB unit value.
• M9 and Above CCL Penetration Rate41% in 2027E and 58% in 2028EThe share was only 7% in 2026E.
• NVIDIA AI Server Rack Forecast92 thousand racks in 2027E; 148 thousand racks in 2028EAI server eight-card-equivalent shipments are expected to be 1.9 million, 2.4 million, and 2.6 million units in 2026E-2028E, respectively.
• Backplane Bull Scenario2028E GPU PCB/CCL market size of US$42.8bn/US$25.9bnAssumes PTFE is used, with backplane PCB/CCL average selling prices 25%/30% higher than the base case.
• Backplane Bear Scenario2028E GPU PCB/CCL market size of US$42.2bn/US$25.3bnAssumes M8 is used, with backplane PCB/CCL average selling prices both 40% lower than the base case; the base-case size is US$42.6bn/US$25.6bn.

Impact & implications

AI infrastructure investment is moving the PCB/CCL industry from pure shipment growth into a phase where specifications, content, and pricing all rise simultaneously. Direct beneficiaries include advanced PCB and CCL manufacturers, while indirect beneficiaries include suppliers of high-end copper foil and low-dielectric glass cloth. Supply-chain competition will shift from ordinary capacity to high layer counts, low-loss materials, customer qualifications, yield, and overseas capacity expansion capabilities. There is uncertainty around the backplane material route, but the impact of different material scenarios on the overall market size is smaller than that of total AI server volume and specification upgrades themselves, so the industry's medium-term direction remains positive.

Risks

• Global AI infrastructure capital expenditure or AI server shipments fall short of expectations.
• Upgrades in AI server PCB layer counts, HDI, and CCL material specifications are slower than expected.
• Certification of M9 quartz-containing glass materials is delayed, or the final material route shifts to lower-value M8.
• Concentrated capacity expansion by major manufacturers causes supply-demand conditions and pricing to be weaker than expected.
• Advanced product yields, customer qualifications, and new capacity ramp-up fall short of expectations.
• Changes in supplier allocation or intensified industry competition lead to share declines for individual companies.
• Exchange rates, raw material prices, and macro demand fluctuations affect the profitability of Japanese and Chinese suppliers.
• High growth expectations have already pushed up target valuations; if earnings delivery falls short of expectations, valuation compression may occur.

What to watch

• Realization of 2027E and 2028E NVIDIA and ASIC AI server rack and eight-card-equivalent server shipments.
• Actual adoption progress and final material route of NVL144 Rubin Ultra backplanes.
• Customer qualification and mass production pace for M9, PTFE, NER Glass, NEZ Glass, and HVLP4 and above copper foil.
• Penetration rates of six-step and above HDI, 30-layer and above PCBs, and M9 and above CCL.
• Allocation changes among PCB and CCL manufacturers at NVIDIA, AMD, AWS, Google, Meta, and Chinese ASIC customers.
• Construction, yield, utilization rates, and mass production timing of new capacity in Thailand, Japan, and China.
• Price increase progress and supply-demand tightness for high-end CCL, copper foil, and glass cloth.