AI Servers Drive PCB Supply Chain Volume and Price Surge; Shengyi's Q3 Net Profit Guidance Beats Expectations by 65%
nashnova research
AI server demand is lifting volume and pricing across all three tiers of China's PCB supply chain. Shengyi Technology's Q3 net profit guidance beat Goldman Sachs' forecast by 65% — from copper-clad laminates to fiberglass cloth to circuit boards, five companies signal the same story: tight supply, rising prices, expanding capacity.
Why did Shengyi's guidance come in so far above expectations?
Q3 net profit guidance of RMB 3.2–3.6 billion, up 212%–252% year-on-year and 49%–68% quarter-on-quarter.
Goldman attributes the beat to three factors happening at once: higher CCL shipment volumes and average selling prices, a richer product mix plus price hikes lifting gross margin, and above-expected non-operating income.
This means → gross margin rose 2.4 percentage points in a single quarter to 35.0%, even as raw materials also climbed. Pricing power outran cost inflation, and the high-end product share grew simultaneously.
Can Q3's profit level be sustained?
Goldman's model projects Q4 net profit falling to RMB 2.148 billion, down 34% quarter-on-quarter, with gross margin retreating to 29.5%.
In plain terms = part of Q3's profit came from non-operating income that may not recur every quarter. Annualizing Q3 would overstate full-year earnings.
Goldman raised Shengyi's 2026–2028 net profit forecasts by 18%, 2%, and 2% respectively, lifted the target price from RMB 249 to RMB 255, and kept the stock on its Conviction Buy list.
Where is the tightest bottleneck upstream?
Kingboard Laminates management pointed to fiberglass cloth as the tightest input. The company raised thin fiberglass cloth prices again in mid-October, after multiple rounds earlier in 2026. AI products require specialty yarns — low-dielectric, low-thermal-expansion, and quartz fiberglass — and the spec upgrade is pushing prices directly.
Some non-AI capacity is shifting to AI products, squeezing ordinary CCL supply as well. This means → AI orders are not only lifting high-end material prices but also indirectly raising ordinary material prices by crowding out capacity.
The core expansion bottleneck is looms: Kingboard has locked in this year's loom supply, but meaningful volume expansion won't arrive until 2027, when major loom suppliers scale up. In plain terms = how fast fiberglass cloth can expand depends on how fast looms can be delivered — 2026 additions are limited.
Why are circuit-board makers diverging?
Shenghong Technology posted Q3 revenue up 26% quarter-on-quarter, but net profit rose only 3%, with gross margin dropping from 32.1% to 30.5%. Goldman attributes this to a platform transition: AI boards are migrating to a next-generation platform, and the changeover temporarily depressed revenue.
Goldman expects Shenghong's Q4 revenue to jump 64% quarter-on-quarter, with gross margin rebounding to 40.9% and net profit surging 143% — driven by the new platform's shipment ramp and higher per-unit pricing.
Shennan Circuits reported first-three-quarter net profit of RMB 4.05–4.15 billion, up 74%–78% year-on-year. UBS estimates Q3 recurring net profit beat its forecast by 21%–28%. Demand spans AI servers, high-speed switches, optical modules, processors, and packaging substrates.
Can this volume-and-pricing logic extend into 2027?
Goldman sizes the global AI-server PCB market at $38 billion in 2027 and $84 billion in 2028, implying a 148% CAGR from 2026 to 2028.
The AI CCL market grows even faster — 161% CAGR over the same period, reaching $48 billion by 2028. This reflects a rising share of higher-priced materials in the mix.
Two key checkpoints ahead: whether loom deliveries arrive on schedule, and whether next-generation AI platform material specs upgrade as expected. A delay in either would undercut the sustainability of this volume-and-pricing cycle.
市场有风险，内容仅供研究参考，不构成投资建议。