Goldman Sachs Significantly Raises AI PCB/CCL Market Size Forecasts, Bullish on Shengyi Tech, Shenghong Tech, WUS Printed Circuit, and Shennan Circuits
Miles Bennett
Goldman Sachs raised its 2027 global AI-server PCB/CCL market forecasts to $38 bn / $22 bn, up 38%/18% from prior estimates, and issued its first 2028 call — $84 bn / $48 bn — driven by shipment volume and rising average prices moving in tandem.
How much bigger did the forecast get?
Goldman lifted its 2027 AI PCB forecast from $27 bn to $38 bn and CCL from $19 bn to $22 bn.
2028 is a brand-new forecast year: PCB at $84 bn, CCL at $48 bn — more than doubling in two years.
The 2026-2028 compound growth rate: roughly 148% for PCB, 161% for CCL. This means → Goldman sees AI PCB/CCL not as a steady-growth segment but as a market in exponential breakout.
What is driving the growth — volume, price, or both?
Volume: PCB shipment area grows at an 85% CAGR through 2028, reaching 4.5 million sqm; CCL shipments grow at 77% CAGR to 131 million sheets.
Price: PCB/CCL average selling prices rise at 34%/48% CAGRs. In plain terms = not only are more boards shipping — each board is getting more expensive.
The price increase comes from spec upgrades: M9 laminate — a higher-performance copper-clad base material — plus 7-/8-layer HDI (high-density interconnect boards) and 30+ layer PCBs. More layers and better materials push unit prices up structurally.
What changes with Nvidia's next-gen architecture?
Goldman raised its Nvidia-driven AI server rack forecast to 92,000 units in 2027 and 148,000 in 2028.
The model now includes Rubin Ultra (NVL72, NVL144, NVL576 configurations); NVL144 is expected to use a backplane design. This means → PCB content per rack rises further, directly expanding the addressable market.
Backplane material remains uncertain. Goldman ran three scenarios: base case on M9, bull case on PTFE — a fluoropolymer with better high-frequency performance but higher cost — and bear case on M8. The 2028 market size gap across all three is modest. This reflects that the material choice is manageable; the real variable is how many racks ship.
Which companies does Goldman favor most?
Shengyi Technology: 12-month target price RMB 247, implying 50.4x 2027E P/E. Net profit forecast to grow 104%/73% in 2026/2027; AI-server revenue share reaching ~40% in 2027, over 50% longer term; operating margin rising from 15% to 20%.
Shenghong Technology: target price RMB 550, implying 26.3x 2027E P/E. Key drivers include PCB layer-count upgrades, PCB replacing copper cables, customer expansion into ASIC — custom AI chip — server PCBs, and Thailand capacity diversification. Its Thailand A1 plant is in mass production; A2 targets Q3 2026 ramp, with A2/A3 output 40%-100% higher than A1.
The buy list also includes WUS Printed Circuit, Shennan Circuits, Panasonic HD, Mitsui Kinzoku, Nittobo, GCE, EMC, Zhen Ding Technology, and TUC.
What is the biggest risk to this forecast?
Goldman flags three: AI infrastructure spend falling short, supply allocation missing expectations, and a shift in the technology roadmap.
In plain terms = if hyperscalers cut AI capex or the next-gen architecture takes a different technical path, those numbers shrink.
One more critical checkpoint: major manufacturers are scaling capacity aggressively. Whether utilization rates hold up after that expansion will determine if the forecasts materialize. This means → supply is ramping — demand must keep pace, or the volume-plus-price thesis breaks.
Content is for reference only, not financial advice.