Citi's Comparative Review of Top 3 AI PCB Players: WUS Hits Record Gross Margin, Shenghong Front-Loads Costs, Shennan's Substrate Business Recovers
Nashnova编辑部
Citi compared H1 results across WUS, Shenghong, and Shennan — all three AI PCB leaders posted strong revenue growth, but margins diverged sharply. The industry has moved from 'who has AI orders' to 'who can turn orders into profit.'
Same AI PCB boom — why are profit pictures so different?
WUS posted Q2 revenue of RMB 7.475 bn, up 20% q/q, beating Citi and Bloomberg consensus by 13–14%. Gross margin rose to 41.3%, a record high.
Shenghong's Q2 revenue hit RMB 6.11 bn, up 29% y/y but 6% below Bloomberg consensus. Gross margin was 32.1%, down 2.4 pp q/q. Adjusted net profit of ~RMB 1.2 bn missed some buyside expectations of ~RMB 1.5 bn.
Shennan's H1 revenue reached RMB 15.3 bn, up 46.3% y/y. The biggest surprise: substrate gross margin jumped to 31.4%, up 16.2 pp y/y.
What's behind the margin divergence?
WUS's story is product-mix upgrade: 800G switch PCBs (printed circuit boards — the "highway" carrying signals between chips) are ramping. Telecom PCB gross margin hit 43.3% in H1. This means → the higher the share of premium products in mature mass production, the more incremental revenue drops straight to profit.
Shenghong's margin dip is not a demand problem — it is cost front-loading. The company hired and trained staff early, and new plants began depreciating before revenue ramped. In plain terms = the factory is built, the people are in place, but the lines are not yet running full — costs arrived before revenue.
Shennan's repair has two layers: PCB and EMS margins improved via mix optimization and price pass-through. Substrate margin — the packaging base that carries high-end chips — leapt from ~15% to 31.4% as BT substrate utilization rose from 80–85% to above 95%, spreading fixed costs thinner.
Can you directly compare their "AI revenue"?
No. The definitions differ widely. WUS gives the clearest breakdown: high-speed switch PCB is 52% of total revenue, AI server and HPC PCB is 18% — together 70%.
Shennan's management says AI PCB is roughly half of PCB sales, and PCB is 56% of group revenue — implying AI PCB is about 28% of total revenue, excluding substrates and AI power products that also benefit.
Shenghong has not disclosed a precise AI share. Management says non-AI revenue will be broadly stable in 2026 and may stay at "several tens of billions" in 2027, with incremental growth driven by AI products.
Raw-material costs are rising — who can pass them on?
All three face cost pressure, but pricing power varies sharply.
Shennan's approach: raise the AI product mix and decline orders that refuse price increases. PCB and EMS margins still hit records. This means → Shennan is actively screening order quality, not chasing volume.
WUS shows the split inside a single company: auto PCB margin fell to 19.0% (down 6.2 pp y/y), while switch products held high margins thanks to technical barriers. In plain terms = within the same firm, products that can reprice and products that cannot are pulling further and further apart.
Where does each company stand on capacity?
WUS is furthest ahead. Its Thailand plant contributed ~RMB 375 mn revenue and ~RMB 43 mn net profit in Q2. Datacom customer certifications are over 90% complete, utilization is at full capacity, and the plant mass-produces up to 30-layer 400G PCBs. 800G sampling has begun.
Shenghong is still in heavy build-out. 2026 capex is capped at RMB 18 bn, with four domestic plants at different stages (ramp, trial production, civil works). Thailand is expected to ramp from late 2026 into 2027.
Shennan's focus is substrates. Its Guangzhou BT and FC-BGA (fan-out chip-scale packaging substrate for high-end chip packaging) project totals ~RMB 6 bn across two phases, with ~RMB 4 bn already spent. FC-BGA revenue grew from RMB 100 mn in 2025 to ~RMB 200 mn in H1 2026. Citi expects Phase 1 FC-BGA to turn profitable in H2 2026 — roughly a year ahead of the original plan.
All rated Buy — so why are target prices moving in different directions?
Shenghong: margins missed; 2026–28 net profit forecasts cut 4–6%; target price lowered from RMB 456 to RMB 345; valuation multiple dropped from 25× to 20× on 2027 earnings.
WUS: earnings beat; 2026 net profit forecast raised 5% to RMB 7.032 bn — yet the target price still fell from RMB 189 to RMB 161. Citi cut the P/E multiple from 30× to 25×. This reflects a cooling in overall AI-sector sentiment: even a strong beat earns a lower premium than before.
Shennan is the only name getting a double upgrade: 2026–28 earnings forecasts raised 11–32% and the valuation multiple lifted too. In plain terms = all three are Buy, but Citi's subtext differs — WUS has proven the profit flywheel works, Shenghong still needs subsequent quarters to show margins can recover, and Shennan's key variable is whether FC-BGA can turn from a loss-making line into a replicable profit center.
市场有风险，内容仅供研究参考，不构成投资建议。