←News detail

Semiconductor Industry ObserverUpdated Oct 30 14:13

Yole Report: AI Is Splitting the PCB Industry into Two Paths: High Growth and Stagnation

AI summary

A Yole Group report shows that in 2025 the global PCB industry's revenue was about US$88 billion, with an overall compound annual growth rate of 8.1%, but the AI-related segment's compound annual growth rate reached 45%–70%, while general-purpose boards were nearly stagnant. By 2031, circuit boards with 18 or more layers will grow at an annual rate of 15.7%, while 4–6 layer boards will grow at only 2.3%; servers and data centers' share of PCB demand has already risen from 16% in 2020 to 25%. Copper-clad laminates iterate one generation every 18–24 months; the Vera Rubin platform uses M8 laminate, and the unit price of M9 reaches four times that of M7.

Body · AI translation

AI is dividing the PCB industry, forming two major route sources: Semiconductor Industry Vertical and Horizontal Published: 2026-10-29 17:42 PCB market data generated poster The AI-related segment's compound annual growth rate can reach 45%~70%, while the ordinary general-purpose board segment's growth is nearly stagnant.

In 2025, the printed circuit board (PCB) industry's revenue was approximately US$88 billion. Conservative forecasts show that, calculated at a compound annual growth rate of 8.1%, the market size will reach US$140 billion by 2031. Behind this overall average lie two sub-tracks heading in completely different directions: the AI-related segment's compound annual growth rate can reach 45%~70%, while the ordinary general-purpose board segment's growth is nearly stagnant.

Against this backdrop, Yole Group released a new market and technology analysis report — the "2026 Printed Circuit Board (PCB) Industry Report." This report covers the complete industry chain: from copper foil, copper-clad laminate (CCL), and glass fiber cloth, to PCB boards, IC substrates, and production equipment, and then to end demand.

The hot directions in the AI infrastructure track are familiar to everyone: 3nm and 2nm logic chips, CoWoS, advanced packaging bottlenecks, co-packaged optics, and the iteration from HBM3 to HBM4E. These are all technologies that the capital market focuses on and that companies prioritize in their layouts, and they are almost all concentrated inside packaging. But current industry discussion ignores the underlying foundational layer. In fact, the PCB and its supporting materials ecosystem are simultaneously undergoing a reconstruction of the same magnitude, and its supply chain structure is even more fragile than that of front-end chip manufacturing. The industry is all focusing on the chip die, yet ignoring the circuit board that carries the chip.

Ignoring this will lead to misjudgments. If the industry's average growth rate of 8% is used to evaluate the PCB market, it is equivalent to placing AI high-end boards growing 20-fold and general-purpose boards with sluggish growth under the same set of assumptions, failing to see the direction of value transfer: value is flowing to high-end boards with 18 or more layers, AI-grade high-speed materials, and new types of circuit boards that did not exist in previous-generation products.

This divergence is reflected in every link of the industry chain. By 2031, circuit boards with 18 or more layers will grow at an annual rate of 15.7%, while 4–6 layer boards will grow at only 2.3%; at the same time, since 2020, the share of servers and data centers in PCB demand has risen from 16% to 25%, replacing consumer electronics as the main demand driver. By 2028, the PCB-related revenue of manufacturers leading the AI supply chain will approach US$109 billion, of which core AI-related business will be about US$76 billion, with an annual growth rate of 76% from 2025 to 2028. In 2025, these leading manufacturers' production capacity will account for about 45% of the global total, and in 2026 their share will exceed half.

Dr. Bilal Hachemi, senior technology and market analyst at Yole Group, said, "The industry average growth rate is the least valuable data point in this market. Half of the track grows at 45%~70% annually, while the other half barely grows. Looking only at the average will lead to wrong judgments about both segments."

Growth is driven by the increase in value per system, not by shipment volume. The PCB value corresponding to the new generation of NVIDIA racks (including the Vera Rubin platform) has increased by more than 200%, of which 57% of the increment comes from new board types that did not previously exist. The PCB value of a custom accelerator is about twice that of a GPU; by 2026, ASICs will account for about 30% of AI servers. The upstream materials market will double to US$46 billion, and AI-specific PCB materials will gradually replace traditional materials.

Automotive electronics is the second-largest growth mainline, and its growth is more sustainable; by 2031, the market size will exceed US$23 billion, with a compound growth rate of 14%. The growth comes entirely from structural changes: circuit boards for electric vehicle powertrains will grow at an annual rate of 30%, and advanced driver assistance (ADAS) central computing boards will grow at 46%; while PCB demand related to traditional fuel vehicles will shrink by 15% per year. After cloud providers' capital expenditure returns to normal, the automotive sector will continue to support demand for high-end HDI boards.

Behind the data, the definition of the circuit board itself is changing. The iteration pace of copper-clad laminates has already formed a cycle like that of chips, with a new generation every 18~24 months: GB300 uses M7 materials, the Vera Rubin platform uses M8, Rubin Ultra in the second half of 2027 will use M9, and the Feynman platform, aimed at 448Gbps signal transmission, will use M10 materials. Each generation upgrade replaces all four main materials; the unit price of M9 materials reaches four times that of M7.

The circuit board is taking on the function of packaging. At a rate of 224Gbps, the skin depth is only 0.27 micrometers. Copper surface roughness, which could be ignored in the previous generation, now directly causes signal loss. This is also why Yole Group previously proposed in its report the semiconductorization of PCBs.

This article is reproduced from media reports or online platforms and represents the author's personal stance or views. Our reproduction is solely for sharing and does not mean that we approve of or agree with it. If the source attribution is incorrect or your lawful rights and interests are infringed, please contact customer service promptly, and we, as a neutral platform service provider, will promptly correct, delete, or handle it in accordance with the law. Comments No user comments yet Latest News More Related Recommendations

Computing power race, power supply looks to GaN 2026-10-30 12:07

Intel appears at Tencent Global Digital Ecosystem Summit, multi-dimensional innovation accelerates agent implementation 2026-10-29 22:43

Physics-based verification, the cornerstone of trustworthy agent EDA 2026-10-29 18:24

Source

Semiconductor Industry Observer