Goldman Sachs expects the AI server PCB market to reach $37.5 billion by 2027, a 38 percent upward revision implying 148 percent annual growth.
Quay lại
Goldman Sachs lifts AI server PCB forecast 38% to $37.5 billion
Goldman Sachs expects the AI server PCB market to reach $37.5 billion by 2027, a 38 percent upward revision implying 148 percent annual growth.
Goldman Sachs sharply raised its forecasts for AI server printed circuit boards and copper-clad laminates, projecting the PCB market to reach $37.5 billion in 2027 and $84 billion in 2028, with growth driven by cloud vendors' in-house ASIC chips as much as Nvidia's GPUs.
"The upward revision is not entirely driven by Nvidia GPU servers — demand growth from self-developed ASIC servers by cloud computing vendors has become a more important incremental driver," the bank said in its global PCB and CCL industry report.
The bank lifted its 2026 global AI PCB forecast 35 percent to $13.56 billion, raising the GPU server segment just 5 percent to $4.728 billion while boosting the ASIC server segment 60 percent to $8.832 billion. By 2027, GPU and ASIC server PCBs are projected at $18.803 billion and $18.726 billion, up 18 percent and 67 percent respectively. The CCL market is expected to reach $22.1 billion in 2027, an 18 percent upward revision, and $48 billion in 2028, implying a 161 percent CAGR.
The forecast revisions point to sustained pricing power for high-end PCB and CCL makers. Hong Kong-listed suppliers rallied on the news, with Shenghong Technology (02476.HK) up 10.91 percent, Kingboard Laminates (01888.HK) gaining 9.91 percent and Delton (01989.HK) climbing 10.93 percent.
ASIC servers drive the bulk of the upgrade
The magnitude of the forecast revisions shows the growth is not solely a GPU story. Goldman raised its 2026 global AI PCB forecast 35 percent to $13.56 billion, with the GPU server segment revised up just 5 percent to $4.728 billion while the ASIC server segment jumped 60 percent to $8.832 billion. By 2027, GPU and ASIC server PCBs are projected at $18.803 billion and $18.726 billion, up 18 percent and 67 percent respectively.
The CCL market shows the same pattern. Goldman raised its 2026 AI CCL forecast 20 percent to $6.956 billion, with the ASIC segment up 36 percent to $4.181 billion. The 2027 market is projected at $22.066 billion, of which ASIC server CCL accounts for $10.061 billion — a 44 percent upward revision.
This shift indicates AI computing infrastructure is diversifying from a GPU-centric platform toward a mix of GPUs, cloud vendors' proprietary chips and other ASIC platforms. For the PCB supply chain, that means a broader customer base, and manufacturers able to serve both GPU and ASIC ecosystems stand to gain more growth.
Volume and prices rise together
PCB and CCL demand is growing on both shipment volume and average selling price. Goldman expects AI server PCB shipment area to rise from 900,000 square meters in 2025 to 4.5 million in 2028, an 85 percent CAGR, while AI CCL demand climbs from 27 million sheets to 131 million, a 77 percent CAGR.
Average selling prices are rising faster. The blended AI PCB ASP is projected to climb from $5,833 per square meter in 2025 to $18,549 in 2028, a 34 percent CAGR, while AI CCL ASP rises from $93 per sheet to $362, a 48 percent CAGR. Simultaneous growth in volume and price is why the market is expanding far faster than server shipments.
Specification upgrades are the core driver of value growth. The share of PCBs with more than 30 layers is expected to rise from 12 percent in 2025 to 47 percent in 2028, while M9-grade and higher CCL materials grow from 7 percent of the market in 2026 to 58 percent in 2028. M9 materials reduce signal loss at high speeds while improving heat dissipation, making them essential as GPU and ASIC platforms push toward higher communication speeds and power consumption.
Tight supply in high-end segments
Despite large-scale capacity expansion across the supply chain, Goldman expects industry capacity utilization to remain elevated over the next two years. High-layer-count PCBs and advanced HDI boards require more equipment time and labor per unit and yield lower returns than standard products, so new nominal capacity cannot be converted into high-end capacity at the same proportion. Customer qualification, equipment commissioning and production ramp-up also take time.
That means high-end segments — PCBs with more than 30 layers, M9 CCL, HVLP4 and above copper foil, and low-dielectric fiberglass cloth — may continue to face tight supply-demand conditions and pricing upside. Competition shifts from output expansion toward yield rates, technical certification, customer share and high-end material supply capability.
For investors, the tight supply points to margin expansion for suppliers with advanced manufacturing. Kingboard Laminates, the global CCL leader, has issued multiple price-increase notices to pass on costs and restore gross margin, while Shenghong Technology, a core global supplier of PCBs for AI computing cards, is positioned to benefit from the shift toward higher-layer-count boards. The bank's base case assumes Nvidia's next-generation NVL144 Rubin Ultra platform will adopt a backplane design using M9 material, with related chip shipments reaching 2,000 units in 2027 and 8,000 in 2028.
This article is for informational purposes only and does not constitute investment advice.
[1] ＜Research＞ G Sachs Sharply Raises PCB and CCL Market Size Forecasts, Expects 2026-28 CAGR at 148% and 161%; Sector to Benefit from Rising Volume and Prices[2] Hong Kong-listed PCB-related stocks continued to rebound, with Goldman Sachs significantly raising its forecasts for AI server PCBs and copper-clad laminates (CCL).[3] Goldman Sachs significantly raised its forecasts for AI server PCBs and CCLs, projecting the market sizes to reach $84 billion and $48 billion, respectively, by 2028.[4] Shares of Kingboard Laminates and two other companies surged more than 14% intraday, driven by strong AI hardware demand and the shift toward optical interconnects boosting related stocks.[5] Hong Kong Market Close | The three major indices diverged this week, with AI-related stocks emerging as the primary focus of investor capital.
