Rating: Cautiously Bullish | Target Price Range: RMB 114–140 | Current Price: RMB 107.06 (2026-10-23 close) | Margin of Safety: +6% (vs. lower bound of base-case fair range) | Time Horizon: 12 months
Currency: Chinese Yuan (CNY). Valuation, market cap, target price, and three-scenario ranges are all computed with the 2026-10-23 closing price of RMB 107.06 as the frozen anchor; total share count 1.924 billion shares, corresponding to a market cap of approximately RMB 206 billion.
WUS Printed Circuit is a global top-tier supplier of high multilayer PCBs for AI data centers (10.3% share in data center boards per CIC, ranked No.1 globally), currently in an accelerating earnings-delivery phase: the 2026 first three quarters preliminary results guidance indicates net profit attributable to parent of RMB 5.45–5.70 billion, up +100.5% to +109.7% YoY; data communications board revenue in 2026H1 was RMB 11.330 billion (+73.5%), gross margin 43.26%. At the current price of RMB 107.06, the stock trades at only 18.7x consensus 2027 net profit (approx. RMB 11.0 billion); reverse DCF shows the market is paying almost nothing for growth beyond 2028—relative to consensus 2027 CSP capex growth (+30% to +38%) and the starting point of 1.6T switch volume ramp, implied expectations look conservative, with valuation reasonably cheap but not deeply undervalued. Key constraints: an industry-wide capacity expansion wave exceeding RMB 80 billion concentrated in 2027–2028, duration divergence over decelerating AI capex growth (Goldman Sachs expects growth to slow to +12% in 2028), the company's own free cash flow turning negative during its expansion phase (-RMB 2.45 billion in 2026H1), and potential dilution from the pending H-share issuance. Base-case fair range: RMB 114–140, probability-weighted expectation approx. RMB 127 (approx. +18.5% vs. current price); bear case RMB 55–82 (-49% to -24%). Odds are skewed positive but not dramatically, hence cautiously bullish rather than aggressively bullish.
Changes underway (next 6–18 months): ① Thailand facility turning from ramp-up drag into profit contributor—2026Q2 single-quarter revenue approx. RMB 375 million, net profit approx. RMB 43 million, achieving a single-quarter turnaround; overall profitability in the first three quarters, with expanded equipment entering a 6–12 month release window after arrival in 2026Q2; ② 1.6T switch PCBs moving from small batches to the main volume ramp in 2027; 224Gbps products already in small-batch supply, with pre-research on 448Gbps and 3.2T supporting products underway—a clear generational product relay; ③ upgraded shareholder-return framework—the controlling shareholder proposed a 2026 cash dividend of no less than RMB 2.2 billion (approx. RMB 960 million/year average over the prior three years), accompanied by a 2026–2028 dividend plan; ④ H-share issuance (refiled 2026-06-05), if completed, would open overseas expansion financing channels and introduce international investors.
Primary stock price driver: Total North American CSP capex and 2027 guidance—the top five customers account for 53.32% of revenue, with the largest at 14.26%; the company is deeply tied to leading cloud providers and equipment makers, and AI compute capex is the master valve of order visibility. Second is single-quarter gross margin pass-through capability amid eight rounds of upstream copper-clad laminate price hikes (2026H1 overall gross margin of 38.72% still hitting new highs), plus the pace of Thailand expansion and new Qing-Song capacity deployment.
Verifiable expectation gap: Sell-side consensus (2026-10-23 snapshot, n=17) for 2026E net profit is approx. RMB 6.62 billion, lagging 10–15% behind the annualized level implied by the Q3 guidance (approx. RMB 7.4 billion)—the guidance shows RMB 5.45–5.7 billion already achieved in the first three quarters, implying Q4 alone needs only RMB 0.9–1.2 billion to reach the consensus midpoint. If the 10-28 Q3 report confirms Q3 net profit within RMB 2.527–2.777 billion with stable gross margins, there is clear room for upward consensus revision. Reverse DCF (WACC 9%) shows the current price implies steady-state free cash flow of only approx. RMB 10.1–12.1 billion/year from 2028 onward, near-zero perpetual growth—an implied assumption contradicting the industry evidence of 1.6T volume ramp and Rubin-generation per-cabinet PCB value leaps. However, it must be emphasized: with 39 sell-side Buys / 0 Sells and a median target price of RMB 161.8 above the current price, "undervaluation" is not the market consensus view; the expectation gap mostly reflects "consensus lagging disclosed results" rather than a broad market misjudgment. Moreover, the stock fell approx. 6.9% cumulatively in the 6 trading days after the Q3 guidance release—the market has already begun digesting crowding by "selling on good news."
Verification and falsification: Upside verification—2026-10-28 Q3 report (Q3 net profit, gross margin, inventory/contract liabilities); North American cloud providers' 2027 capex guidance in Jan–Feb 2027 (consensus approx. USD 1.2–1.3 trillion, +30% to +38%); sequentially expanding single-quarter profits at the Thailand subsidiary. Falsification conditions—2027 aggregate CSP capex guidance growth falling below 20%; single-quarter gross margin declining QoQ for two consecutive quarters and falling below 36%; Q3 net profit below the guidance floor of RMB 2.527 billion.
Key evidence:
Quantified demand-side chain: The four major North American cloud providers' 2026 capex approx. USD 602–725 billion (vs. approx. USD 410 billion in 2025); on the switch side, 800G ports exceeded 25% of data center shipments in 2025, with 1.6T shipping in small batches in 2026 and entering the volume ramp in 2027; AI server accelerator boards rise from 8–12 layers in general servers to 18–22 layer high-end HDI with M8-grade materials, at unit value multiples of traditional boards. Sell-side estimates (Soochow Securities 2025-09, cited) put Rubin-generation rack PCB value approx. +179% vs. GB200; subsequent BOM teardowns (Sinolink 2026-05, BofA 2026-09) revised this up to approx. +233% (baseline adjusted to GB300)—generational estimates fluctuate widely before mass production; the direction (large leap in unit value) is more reliable than any single figure. The company's positioning: 10.3% global share in data center PCBs, ranked No.1 (CIC, revenue for the 18 months to 2025-06); No.1 globally in 22-layer-and-above PCBs; 2026H1 data communications growth (+73.5%) outpacing Prismark's server/storage segment growth (2025: +43.6%), implying continued share gains.
Two boundaries must be objectively noted: First, the 82.8% datacom revenue concentration is a double-edged sword—it is the source of this round's earnings acceleration and also the transmission coefficient of risk when AI capex recedes. Second, there is genuine institutional divergence on the "continuation into 2028" duration judgment: Goldman Sachs' strategy team (2026-09-25) expects hyperscaler capex growth to slow from approx. +100% in 2026 to +54% in 2027 and +12% in 2028, while JPMorgan (2026-10) raised its 2027 data center capex growth forecast to +70%, and Morgan Stanley's measure (incl. leasing) puts 2027 at approx. USD 1.58 trillion—a forecast dispersion band from +12% to +70%. We treat 2027 as the primary holding verification period, with 2028 duration as a scenario variable rather than a baseline assumption.
Key evidence:
Earnings quality has multiple cross-validations: the gap between non-GAAP and reported profit is small (2026H1 non-recurring gains RMB 129 million, 4.4% of total); 2025 OCF/net profit at 1.01x—profits are cash-backed; ROIC (net profit / (equity + interest-bearing debt)) estimated at approx. 19% in 2025 and annualized approx. 22% in 2026H1, significantly above the approx. 9% cost of capital—this is a genuine moat, not fake growth. But the episodic weakening of 2026H1 cash metrics must be faced squarely: cash-to-profit ratio (OCF/net profit) fell to 0.39 (1.25 in the prior-year period); inventory rose to RMB 6.041 billion (+42.3% from year start); FCF was -RMB 2.451 billion; interest-bearing debt rose from RMB 4.56 billion to RMB 8.31 billion (+82.5%); debt-to-asset ratio 52.2% (38.7% at end-2023). The company's explanation is consistent: stockpiling (locking in materials amid the copper-clad laminate price-hike cycle) plus capacity expansion (Thailand expansion equipment arrival; three major projects incl. Qing-Song totaling approx. RMB 17.6 billion in construction). This is a deliberate choice to "convert profits into capacity" rather than deteriorating receivables, but it holds only if demand persists after capacity ramps—if 2027–2028 conditions weaken, inventory write-downs and new depreciation (approx. RMB 1.2 billion/year once the three projects are fully transferred to fixed assets) will amplify profit volatility in reverse.
Key evidence:
Reverse DCF (WACC 9%, perpetual growth 3–4%) shows: the RMB 206 billion market cap implies steady-state FCF of approx. RMB 10.1–12.1 billion/year from 2028 onward—roughly equal to 2027 consensus net profit, i.e., the market is pricing in "2027 as peak earnings, then a plateau." This contradicts demand-side evidence (2027 CSP capex consensus +30% to +38%; 1.6T only entering volume ramp in 2027; data center PCB being the fastest-growing downstream segment industry-wide); under a "2027 earnings peak" reading, it instead suggests the market has already priced in a conservative discount for AI capex duration.
Composition of the base-case fair range of RMB 114–140: 2027E net profit attributable to parent of RMB 11.5–12.8 billion × exit P/E of 19–21x. Anchors for the exit multiple: Victory Giant Technology (the other AI-board twin) currently at approx. 21.6x 2026E; Goldman Sachs' 2026-10-12 target price of RMB 148 for WUS based on approx. 23x 2027E; Shennan Circuits implying 22–25x 2027E—we take 19–21x, below these anchors, reflecting two conservative adjustments: the AI-chain market compresses multiples on forward earnings faster than on current-year earnings (under a more aggressive forward framework—if the market applies only approx. 12–15x to the consensus-extrapolated 2027E net profit for Victory Giant, WUS's fair value would drop to approx. RMB 98–115, near the current price); and the 39 Buy / 0 Sell one-sided expectation structure is itself a source of fragility. Sensitivity must be stated explicitly: this thesis depends on a 19–21x exit multiple; if the market shifts to 12–16x forward pricing, the current price is already close to fair and the upside disappears. This is the lowest-confidence thesis in this report.
Key evidence:
Supply-side math: New capacity from all players concentrated in 2026H2–2028; the company's own three projects, once fully transferred to fixed assets, add approx. RMB 1.2 billion/year in depreciation (approx. 3.9pp of gross margin on RMB 32 billion of revenue); upstream copper-clad laminate has risen approx. +50% cumulatively over eight rounds of hikes this year, with raw materials exceeding 50% of cost of sales. If AI capex growth decelerates significantly in 2027 (Goldman's measure: +12% in 2028) and overlaps with new capacity release, datacom boards could replay the automotive boards' transmission failure: using the automotive board's -6.22pp gross margin sample, a 4–5pp pullback in datacom gross margin would cut 2027E net profit to the RMB 7.0–8.5 billion range; at stress-period valuations of 15–18.5x (referencing the measured cycle-bottom of approx. 14.8x in January 2023 and approx. 17.2x at the September 2024 inter-cycle trough), this corresponds to a bear-case fair value of RMB 55–82, -49% to -24% vs. the current price. A deeper cycle-bottom scenario (30% gross margin, RMB 30 billion revenue, approx. RMB 3.8 billion net profit × 15–16x, approx. RMB 30–48) belongs to a multiple-deterioration scenario for 2028 and beyond—a stress reference rather than a 12-month baseline. Current judgment: high-end segments (M8-and-above materials, 38-layer-and-above processes) remain a seller's market through 2026–2027; price wars are more likely to start in mid/low-end tiers; a substantial portion of the RMB 80 billion expansion figure is indicative investment and not in the highest-entry-barrier segments—the supply risk is real, but the extent of market discussion may slightly front-run the actual delivery pace.
Key evidence:
Three details must be presented faithfully: First, the RMB 2.2 billion dividend is a controlling shareholder proposal, not an adopted resolution; preconditions include "no major investment or major cash outlay needs beyond established plans"—and the three-year dividend plan lists "no major outbound investment plans within the next twelve months" (threshold approx. 30% of audited net assets, i.e., approx. RMB 5.4 billion) as a cash dividend condition—the company has already announced approx. RMB 17.6 billion in projects this year with 2026 capex annualizing at approx. RMB 7.2 billion, so the dividend commitment and investment plans are in tension at the terms level; the final outcome rests with the March–April 2027 annual report proposal and shareholder meeting deliberation. Second, H-share issuance size has no official disclosure; the rumored USD 1.5 billion at a moderate discount implies approx. 5%–7% dilution; higher dilution (9%–11% in some sell-side scenarios) would require an issuance of USD 1.9–3.3 billion—no evidence of that currently. Third, the "H-share discount dragging down the A-share valuation anchor" mechanism has not generally played out in recent A+H samples (CATL's H shares trade at a 29%–40% premium to A shares; several hard-tech A+H names show H>A), but international comparables for the PCB sector trade below battery leaders, so a discount scenario cannot be ruled out. On balance: H-share financing supporting overseas expansion and introducing international investors is a neutral-to-positive event; the dividend proposal's signaling value (a significant increase from the prior three years' RMB 0.96 billion/year, timed closely with H-share progress and carrying pricing-endorsement intent) outweighs the dividend yield itself (RMB 2.2 billion implies approx. 1.07%).
| Metric | 2023FY | 2024FY | 2025FY | 2026Q1 | 2026H1 (Latest) |
|---|---|---|---|---|---|
| Operating revenue (RMB 100mn) | 89.38 | 133.42 | 189.45 | 62.14 | 136.89 |
| Revenue YoY | — | +49.3% | +42.0% | +53.9% | +61.2% |
| Net profit attributable to parent (RMB 100mn) | 15.13 | 25.87 | 38.22 | 12.42 | 29.23 |
| Net profit YoY | — | +71.1% | +47.7% | +62.9% | +73.7% |
| Non-GAAP net profit attributable to parent (RMB 100mn) | 14.06* | 25.46 | 37.61 | — | 27.95 |
| Gross margin | 31.0% | 34.5% | 35.4% | — | 38.7% |
| Net margin (attributable to parent) | 16.9% | 19.4% | 20.2% | 20.0% | 21.4% |
| Operating cash flow (RMB 100mn) | 22.43 | 23.25 | 38.72 | — | 11.51 |
| Free cash flow (RMB 100mn, OCF−CapEx) | 14.33 | 1.77 | 6.74 | — | -24.51 |
| Cash + time deposits due within one year (RMB 100mn) | — | — | 25.79 (cash) | — | 66.65 |
| Interest-bearing debt (RMB 100mn) | 36.27 | 45.56 | 45.56 | — | 83.15 |
| Debt-to-asset ratio | 38.7% | 43.8% | 46.5% | — | 52.2% |
| Weighted ROE | 18.3% | 24.25% | 28.57% | — | 17.38% (half-year) |
*2023 non-GAAP figure follows the comparative column basis of the 2024 annual report; 2026Q1 non-GAAP and cash flow details are not listed separately above, with direction consistent with H1. Balance-sheet stock items (cash/cash-like/interest-bearing debt) are as of the 2026-06-30 semi-annual report; cash-like assets of RMB 6.665bn = cash of RMB 4.554bn + non-current assets due within one year (time deposits) of RMB 2.112bn, excluding time deposits over one year within other non-current assets not separately disclosed; interest-bearing debt = short-term borrowings 38.68 + long-term borrowings 39.84 + amounts due within one year 4.62 (RMB 100mn), with net debt of ~RMB 1.65bn. During this asset-heavy expansion phase, net debt/EBITDA remains low (2025 EBITDA on the order of RMB 4.5bn; net debt/EBITDA <0.5)—leverage risk is manageable but trending upward.
Reasons for metric changes (company explanations provided for all items with YoY ≥ ±20%):
The latest periodic report is the 2026 semi-annual report (disclosed 2026-08-26); the Q3 report is scheduled for 2026-10-28, but the company issued a nine-month earnings preannouncement on 2026-10-15—currently the most important incremental information.
Key preliminary figures: 2026Q1–Q3 net profit attributable to parent of RMB 5.45–5.70bn, +100.54% to +109.74% YoY; non-GAAP RMB 5.36–5.55bn. Backed out from actual H1 of RMB 2.923bn, Q3 standalone net profit attributable to parent of RMB 2.527–2.777bn, +144% to +168% YoY and +50% to +65% QoQ—growth is not only persisting but accelerating. Reason for results change (company's wording): AI data center demand driving volume growth in high-speed switches and AI server PCBs, plus capacity release and improved operational efficiency; the Thai subsidiary has moved from capacity ramp-up into scaled operations, turning profitable in the first three quarters.
Comparison with prior periods and market expectations: The Q3 standalone preannouncement midpoint (~RMB 2.65bn) is 2.6x Q3 2025 standalone (RMB 1.035bn); the nine-month midpoint of RMB 5.575bn already reaches 84% of the sell-side consensus full-year estimate (RMB 6.62bn)—consensus lags disclosed results significantly. Immediately after the preannouncement, Huachuang Securities raised its 2026–28 net profit forecasts to RMB 8.39/13.96/19.97bn, with 2026E implying a Q4 standalone of ~RMB 2.8bn. Our own forecast (2026E net profit attributable to parent of RMB 7.3–7.8bn) sits between the consensus median and Huachuang: implying a Q4 standalone of RMB 1.75–2.2bn, -12% to +8% QoQ vs Q3—fully reflecting the marginal gross-margin pressure from the cluster of Q4 CCL price increase letters and FX uncertainty.
Interpreting the market reaction: After the preannouncement (after close 10-14/10-15), the stock fell rather than rose—from a 10-14 close of RMB 114.97 to a 10-23 close of RMB 107.06, cumulatively about -6.9%; the PCB sector broadly pulled back from highs over the same period (Shengyi Technology down ~36% YTD, its chairman having proposed a buyback to support the stock). This "sell the news" combination indicates: part of the earnings beat was already priced in (stock up 57% over the prior year, and at one point a 31% drawdown from its 52-week high of RMB 154.99), and with a one-sided structure of 39 buy/0 sell ratings, marginal buyers are exhausted. Three hard validation points for the 10-28 Q3 report: where Q3 standalone actuals land within the range; whether quarterly gross margin holds above 38% (a direct test of pricing pass-through); and the QoQ direction of inventory/contract liabilities/operating cash flow (a forward indicator of order backlog).
Business model in brief: Asset-heavy, build-to-order custom manufacturing. Revenue has no subscription/recurring component—entirely order-based sales; products are 18+ layer high-multilayer boards and high-end HDI, serving switches/routers/AI servers/automotive. Pricing power is structurally tiered: data communications boards hold structural pricing power (segment gross margin rising yearly from 35.85%→39.68%→43.26%; 32-layer-and-above product revenue +190.8% in FY2025); automotive boards have no pricing power (company itself cites "lagged cost pass-through"). This is a business where "technology generation determines bargaining power"—not a homogenized cyclical commodity.
Cash conversion of earnings test: OCF/net profit attributable to parent: 2023 1.48 → 2024 0.90 → 2025 1.01 → 2026H1 0.39; FCF/net profit: 0.95 → 0.07 → 0.18 → -0.84. By strict standards, three consecutive years of FCF/net profit below 0.5 must be flagged: the bulk of book profit has not converted into freely disposable cash but entirely into capacity—cumulative capex since 2024 of ~RMB 9.1bn (cash basis), against net profit attributable to parent of ~RMB 9.3bn over the same period. This is not deteriorating collections (receivables scale matches revenue growth; 2025 OCF/net profit of 1.01) but a deliberate reinvestment choice; the key to judging its quality is incremental return on capital: ROIC of ~19–22% against a ~9% cost of capital, ~RMB 0.21 of capital tied up per RMB 1 of incremental revenue in 2025, and projects under construction (Thailand expansion already profitable; the RMB 4.3bn Qing-Song project aimed at AI support) with clear order pull—current evidence supports "high-return reinvestment" rather than a "capital black hole," but the 2026H1 cash conversion ratio of 0.39 and inventory +42.3% need continued verification in the Q3 report and annual report. Also note non-recurring items: 2026H1 non-recurring gains/losses of RMB 129mn, 4.4% of net profit attributable to parent (including RMB 46mn gain on sale of Huangshi supply chain equity)—the small gap between non-GAAP and GAAP shows core earnings are not dressed up by one-off items; under A-share standards there is no large SBC add-back issue.
Return on capital: 2025 weighted ROE of 28.57% (rising yearly from 18.3% in 2023), with estimated ROIC of ~19–22%—significantly above the 15% "true moat" threshold and the cost of capital.
Capex quality: CapEx/depreciation ~5.5x in 2025 and ~9.4x in 2026H1—all expansionary investment (Thailand expansion, Qing-Song, Huli Weidian, Huangshi); maintenance spending not separately disclosed (approximating a lower bound with annual depreciation of RMB 580–760mn). The company does not disclose a maintenance/growth split; judged indirectly via the project list and order pull: expansion aligns with its fastest-growing downstream (AI data center boards).
Moat and quality red flags: Moat = 20+ years of top-tier customer certification barriers (NVIDIA supply chain, North American CSPs, Cisco) + process yield in 38–46 layer high-speed boards (800G yield above 90%) + scarcity of high-end capacity (2026Q1 utilization ~99%). Three quality red flags: persistently negative FCF (as above); rising customer concentration (top five 49.78%→53.32%, largest at 14.26%); large FX exposure with late-start hedging (2026H1 FX loss of RMB 276mn, 9.4% of net profit attributable to parent).
Consistency of words and deeds: pragmatic, with a good track record. Three commitments vs delivery: ① 2025 annual report outlook "Thailand data communications business unit utilization above 90% in 2026Q1, expansion released in 2026Q2" → 2026 semi-annual report confirms full loading, equipment arriving in Q2, Thai subsidiary turning profitable in Q2 standalone (revenue RMB 375mn/net profit RMB 43mn), profitable overall in the first three quarters—delivered and early; ② "fully advancing the H-share issuance" → filed 2025-11-28, updated submission 2026-06-05—in progress but incomplete, slower paced than the "top priority of 2026 key work" wording; ③ FY2025 dividend of RMB 5 per 10 shares, RMB 962mn total—implemented. The company is consistently cautious with forward-looking statements (research meeting minutes repeatedly state they "do not constitute profit forecasts or guidance"), with no record of divergence between preannouncements and formal reports.
Shareholder friendliness: neutral-to-friendly, improving. Absolute dividends flat at ~RMB 960mn for three consecutive years; payout ratio passively diluted from 63.3% to 25.2% (profits surged without a dividend increase—"declining payout ratio"); on 2026-09-30 the company issued a three-year dividend plan (cumulative no less than 30% of average annual distributable profit) + the controlling shareholder proposed in October a 2026 dividend of no less than RMB 2.2bn—direction clearly improving, but terms include investment preconditions (see C5). No buyback record; no equity refinancing history since listing; 2024 option incentive exercise diluted ~0.3%; the H-share issuance is the first substantive dilution point.
Risk signals: No controlling-shareholder reduction record (Bijing Holdings 19.32%, WUS GROUP 11.26%, stable); related-party transactions controlled (annual sales to associate Schweizer of RMB 441mn, 3.22% of same-type transactions, within quota); the only executive reduction in the past year was director/board secretary Li Minggui's 55,000 shares (0.003% of total shares, symbolic). Execution details to track: land for Phase II of the RMB 5.5bn Kunshan project once failed to be delivered on schedule (adjusted to Phase III land in 2026-08)—friction exists in execution cadence of large projects.
| Segment | 2026H1 Revenue (RMB 100mn) | Share | Gross margin | GM YoY | Revenue YoY | One-line business logic |
|---|---|---|---|---|---|---|
| Data communications (incl. AI data centers) | 113.30 | 82.8% | 43.26% | +4.60pp | +73.5% | AI server + switch high-multilayer boards; switches +109.9% in FY2025; AI server/HPC reached RMB 1.828bn in 2026H1 |
| Smart automotive | 14.38 | 10.5% | 18.97% | -6.22pp | +1.1% | Automotive-grade boards; smart/EV subcategories +12.2% are the engine, but lagged cost pass-through squeezes margins |
| Industrial control and others | 2.26 | 1.7% | 40.57% | -4.18pp | +14.1% | Small but high-margin; not the main line |
| Other business (scrap, etc.) | 7.98 in FY2025 | 4.2% (FY2025) | Very thin | — | +60.4% (FY2025) | Scrap recovery sales accompanying scale expansion |
Profit-driven segment: Data communications, with 82.8% revenue share × 43.26% gross margin, contributes ~96% of gross profit (113.30×43.26%≈RMB 4.90bn vs RMB 270mn for automotive boards)—this is in effect an "AI data center PCB single-engine company," with automotive boards having degraded from a second engine to a supplement. A 24.3-percentage-point gross margin gap (data communications 43.26% vs automotive 18.97%): the difference lies not in customer industry per se but in supply-demand and bargaining power—scarce high-end datacom board capacity (dual barriers of certification + process) confers pricing power over downstream, while automotive boards sit downstream of the OEM price war transmission chain with abundant peer capacity. This structure also means: consolidated gross margin (38.7%) is almost entirely determined by datacom boards; automotive drag has been diluted to the margin.
Accounting red flags: ① Large FX gain/loss swings with late-start hedging (hedging only announced 2026-03)—exports account for ~86% of PCB revenue, and FX swings can consume ~9% of net profit attributable to parent in a single period (medium; semi-annual report "YoY changes in key financial data"); ② persistently rapid inventory growth with routine write-downs (inventory +74.3% in FY2025, +42.3% in 2026H1 to RMB 6.041bn; asset impairment losses of RMB 160–270mn for three consecutive years)—growth persistently above revenue; concentrated write-down risk exists if the cycle turns (low; "material changes in asset composition"); ③ related-party transaction scale needs tracking (annual sales of RMB 441mn to 19.74%-owned Schweizer; 2026 quota of RMB 874mn; proportion passively rising as automotive business shrinks) (low; material related-party transactions).
Cross-period consistency: ① The systematic decline in FCF/net profit attributable to parent of 0.95→0.07→0.18→-0.84 is consistent with management's "capacity expansion" explanation (expansion investment disclosed in two consecutive MD&As)—an explainable investment cycle, not a profit-manipulation signal—consistent; ② absolute cash dividends flat for three years with payout ratio passively diluted to 25.2%, and the company only responded with the three-year plan in 2026-09—the company had not explained (now partially addressed); ③ automotive board gross margin declining consecutively 24.45%→22.84%→18.97%, with the company explaining each period "industry competition and raw material cost pressure / lagged cost pass-through"—consistent; ④ cash conversion ratio plunging 1.25→0.39, attributed by the company to stocking purchases—consistent, but requires Q3 report verification of the reversal.
Conclusion: Based on the 2023–2026H1 financials extracted here, no profit-dressing financial engineering was found (small non-GAAP/GAAP gap, R&D fully expensed, no goodwill); the main blemishes are lagged FX hedging management and the pace of inventory expansion—both have company explanations and are trackable.
Current market data (close of 2026-10-23): share price RMB 107.06, market cap RMB 206.0 billion, total shares outstanding 1.924 billion. PE(TTM) 40.7x (84th percentile over the past 5 years), PB(MRQ) 11.45x (89th percentile over the past 5 years) — percentiles are for reference only, not a headline cheapness/expensiveness signal: the TTM denominator (RMB 5.06 billion) badly lags the current situation (Q1–Q3 preannouncements have already reached RMB 5.45–5.7 billion); the understated denominator during the ramp-up phase mechanically inflates PE(TTM); the high PB is supported by a rising ROE (TTM ~30%), implying an expectation of "ROE sustained at 30%+", which depends on the continuation of the AI capex cycle. On a realized basis: the current price corresponds to 27.7x the annualized 2026E preannounced net profit, 18.7x the 2027E consensus of RMB 11.0 billion, and 14.8x Huachuang's RMB 13.96 billion estimate; PEG ~0.4 (linear distortion in the 60%+ growth range, cross-validation only).
Peer benchmarking (snapshot from late October 2026; forward figures are consensus or this report's estimates):
| Company | PE(TTM) | 2026E PE | PB | 2025 growth/ROE | Positioning difference |
|---|---|---|---|---|---|
| Shengyi Technology (WUS) | 40.7 | 27.7 (preannounced annualized basis) | 11.5 | Net profit +47.7%; ROE 28.6% | Global No.1 in data center boards (10.3%), tops in switch segment |
| Victory Giant Technology | 35.9 | 21.6 | 4.8 | Net profit +273.5%; ROE 33.8% | Global No.1 in AI/HPC boards (13.8%), tied to compute cards/UBB, steeper elasticity |
| Shennan Circuits | 38.0 | ~33 | 9.6 | Net profit +74.5%; ROE 20.6% | PCB + substrates + assembly, RMB 17.0bn capacity expansion, tight cash flow |
| Luxshare-precise (DSBG/Dongshan Precision) | 58.4 | ~33 | 8.6 | Net profit rebounded after decline; ROE 6.9% | FPC leader pivoting to AI, sprawling businesses, low earnings purity |
| Avary Holding | 43.2 | ~55 | 4.8 | Net profit +4%; ROE 11.3% | Global No.1 by revenue but consumer-electronics heavy, AI revenue still small |
| Foxconn Industrial Internet (reference) | 16.0 | 16.0 | — | Net profit +74.7% (2026E) | AI server ODM, net margin ~3.5%, multiples not directly comparable |
Market-implied expectations: What must the company deliver to justify RMB 107.06? Reverse DCF (WACC 9%, g 3–4%) shows a market cap of RMB 206.0 billion implies steady-state FCFF of about RMB 10.1–12.1 billion per year from 2028 — roughly the level of 2027E consensus net profit, i.e., the market is paying for "peak earnings in 2027, zero growth thereafter." Reality: 2026Q1–3 has already delivered RMB 5.45–5.7 billion (+100%+); consensus has been revised up 10.1% over 90 days and still lags by 10–15%; 2027 cloud capex consensus is +30% to +38% (JPMorgan +70%), 1.6T is only at the start of its volume ramp, and 3.2T/448G pre-research is underway. The gap between implied expectations and industry evidence is the core basis for this report's bullish stance — but the same math can be read conservatively: if Goldman Sachs' "2028 capex +12%" plays out, "peak in 2027" is precisely the correct pricing, and 18.7x is no longer cheap. Two-sided test conclusion: as long as 2027 capex growth does not fall below +20%, the price-implied expectation is conservative; below that, the current price turns fairly valued to expensive.
Three-tier value (EPV): Asset value ~RMB 9.35/share (net assets attributable to parent ~RMB 18.0bn at end-2026H1 / total shares) — the floor near the far end of replacement cost; EPV zero-growth value ~RMB 34.9/share (normalized FCFF RMB 3.22/share ÷ WACC 9%, minus net debt of ~RMB 0.86/share; normalized basis: 2026 run-rate revenue ~RMB 31.4bn × normalized net margin 22.5%, plus D&A RMB 0.85bn, minus maintenance capex RMB 0.7bn; planned capacity and unfulfilled demand excluded) — only ~33% of the current price is supported by zero-growth cash flow; ~67% is a growth option. That option has real industry support (1.6T ramp, Rubin generation, Thailand ramp-up) and is not pure narrative; but by this report's methodological discipline it must be stated: a high growth-option share neither proves overvaluation nor undervaluation — it hands the judgment to the probability that the ramp scenarios materialize.
Three scenarios and odds (12-month horizon; fair-value ranges consistent with the structured scenarios below):
| Scenario | Probability | 2027E net profit attributable | Exit PE | Fair range | vs. current price |
|---|---|---|---|---|---|
| Bear | 25% | RMB 7.0–8.5bn (capex slowdown + supply release, datacom gross margin down 4–5pp) | 15–18.5x (cyclical-stress valuation, anchored at the 2023 major bottom of 14.8x / 2024 intermission bottom of 17.2x) | RMB 55–82 | -49% to -24% |
| Base | 50% | RMB 11.5–12.8bn (1.6T ramp + Thailand at full capacity, gross margin 36–37%) | 19–21x (anchored at Victory Giant's current 21.6x / Goldman 23x / Shennan implied 22–25x, deliberately haircut) | RMB 114–140 | +6% to +31% |
| Bull | 25% | ~RMB 14.0bn (Huachuang's estimate materializes, 2027 cloud capex revised up to USD 1.3tn+, price increases persist) | 24–27x (growth-continuation pricing, anchored at the lower end of its own 5-year median PE of 30.8x) | RMB 174–196 | +63% to +83% |
Bear-case external cross-check: the most pessimistic sell-side target price findable in the market is RMB 74.92–86.8 (lower bound of Moomoo/Investing.com aggregates, snapshot 2026-10-24) — but its denominator still assumes earnings delivery, making it "the most pessimistic seller in an optimistic market"; our bear case (RMB 55–82) is constructed on independent stress assumptions of earnings downgrades + multiple compression, and being deeper than the market floor is reasonable — we do not adjust for covered sell-side targets. The current price of RMB 107.06 sits in the transition band between the bear-case upper bound and the base-case floor — odds are moderately positive but not lopsided: about +19% to the base-case midpoint, +83% to the bull top, -49% to the bear bottom; upside/downside odds roughly 2:1.
Own earnings forecast (reconciliation anchor): 2026E revenue RMB 30.5–32.0bn, net profit attributable RMB 7.3–7.8bn (Q4 implied RMB 1.75–2.2bn); 2027E revenue RMB 39.5–42.5bn, net profit attributable RMB 10.5–12.8bn. Triangulated against management messaging (no formal guidance; per preannouncements and IR notes), sell-side consensus (2026E RMB 6.62bn / 2027E RMB 11.0bn), and Huachuang (RMB 8.39/13.96bn): we are above the consensus median (which lags) and below Huachuang's upper end (whose 2027E implies a 32.5% net margin requiring gross margins to keep setting records). H-share sensitivity: if 10% new shares are issued at a moderate discount, EPS dilution is ~9%, cutting the base-case fair-value midpoint from ~RMB 120 to ~RMB 109–112.
Conclusion: fairly valued to undervalued; target price RMB 114–140 (base-case fair value); judge quality and price separately. Quality: global No.1 in data center boards, ROE 28.6%, ROIC roughly twice the cost of capital, clear generational positioning in customers and capacity — solid. Price: 18.7x 2027E is low relative to demand-side math, but sensitivity to the exit-multiple assumption is enormous (at 12–16x, fair value is only RMB 98–115), PB at 11.5x sits at the 89th historical percentile, and market expectations are one-sided — this is not a deep margin of safety, but an intermediate state where "continued earnings verification triggers a valuation switch, while an interruption brings a limited but real drawdown." On the anchoring discipline behind the 19–21x base-case exit multiple: the anchors are peer current multiples (Victory Giant 21.6x) and named sell-side forward multiples (Goldman 23x), not the company's own current multiple; a deliberate haircut of ~2–3 turns reflects forward-multiple compression risk and a crowded-consensus discount.
Industry size: The global PCB market was ~USD 69.5bn in 2023 (-15.0%, consumer electronics destocking), ~USD 73.0bn in 2024 (+5.8%), ~USD 85.2bn in 2025 (+15.8%, AI-driven upside; Prismark's April 2026 revised basis), 2026E ~USD 95.8bn (+13%), with a 2025–2030 CAGR of ~7.7% (85.2 → 123.3bn USD; Prismark's 2026 website basis). Mainland China's 2024 output was USD 41.2bn, 56% of the global total, with its share expected at ~60% in 2026. Structural divergence far exceeds the headline total: the server/data storage segment is the industry's growth engine — USD 8.20bn in 2023 (11.8% of total) → USD 10.92bn in 2024 (14.8%) → USD 15.68bn in 2025 (18.4%, +43.6% YoY); Prismark projects a 2025–2030 CAGR of ~15.8% (publicly cited basis; AI server sub-segment ~24.3%); 18+ layer high-multilayer board output value rose +41.7% YoY in 2025. The headline source for industry size figures is Prismark (as cited in Soochow Securities' September 2025 deep dive, Cailianshe, and Securities Times), a publicly cited layer of industry-association-grade primary data.
Value chain and value distribution: Upstream copper-clad laminate (CCL) is the core substrate (cost structure roughly copper foil 40% / resin 26% / glass cloth 18%); the CCL industry is highly oligopolistic — global CR10 reached 77% in 2024 (Kingboard 14.3%, Shengyi Technology 13.6%, EMC 13.3%); further upstream, HVLP4/5 ultra-thin copper foil (monopolized by Mitsui Mining), M8/M9/PTFE resins (Shin-Etsu/Panasonic/EMC), and Low-Dk/quartz glass cloth (Nittobo) are chokepoint segments, with price hikes of 20%–50% since 2025H1. Midstream PCB manufacturing has 2,200+ players globally with CR10 under 40%, but only a dozen or so players at the high-multilayer/HDI high end. The ultimate payers downstream are top North American CSPs and Nvidia-chain equipment makers; order transmission runs CSP capex → system vendors → EMS/ODM → PCB makers. Margins currently sit at both ends: high-end CCL upstream (rising volume and price) and high-end PCB midstream (WUS gross margin 36.9% → 38.7% vs Avary's 21.5%). Upstream, WUS secures priority supply via top-tier customer endorsement (gross margin still +1.06pp during the 2025 CCL price-hike wave); downstream, it retains strong bargaining power via certification barriers and scarce capacity — an advantaged position in a barbell-shaped bargaining structure. Smaller players struggle to pass through CCL price hikes, accelerating their exit and passively raising concentration at the top.
Supply-demand and competitive landscape: Demand side — volume: AI servers and 800G/1.6T switch ramp (Nvidia FY2026 revenue USD 215.9bn, +65%, networking business growth ~+199%; the four big North American cloud capex of USD 602–725bn in 2026, consensus USD 1.2–1.3tn in 2027); price: AI server boards moving from 8–12 layers to 18–22-layer M8 grade (18+ layer unit prices ~3x those of 12–16 layers, per Prismark); switches 800G (12–14 layers) → 1.6T (16–20 layers + M9). Supply side — high-end capacity in tight balance: only a dozen or so players — WUS, Victory Giant, Shennan, Taiwan's Gold Circuit/Unimicron, TTM, etc. — can mass-produce high-end AI server HDI and 800G+ switch PCBs; WUS's 2026Q1 utilization was ~99%; the expansion wave is unprecedented (see risk disclosure in the overall conclusion), but high-barrier capacity (M8 materials / 38+ layer processes / certification) releases concentrated in 2027H2–2028. Entry barriers are fourfold: technology (224G SerDes requires M9/PTFE + 44-layer boards, mSAP line widths of 15–20 microns, long yield ramp cycles), certification (1–2+ years with top customers, high switching costs), capital (a single high-end plant costs RMB 4.0–7.0bn), and environmental/geopolitical factors. Mid-term substitution threat is limited: CPO/glass substrates restructure part of the interconnect architecture, but demand for large-format high-speed PCBs in server/switch motherboards and backplanes is irreplaceable, and CoWoP (chip-to-PCB direct attach) actually raises process and value requirements.
Cycle and regulation: The industry is in a K-shaped split of "AI-driven structural boom + weak consumer electronics recovery," with AI/datacom in mid-upcycle (leading indicators: CCL price-hike rounds, top players' utilization, Nvidia order visibility, and Taiwan-listed monthly revenues all at high levels), but the concentrated supply release in 2027–2028 is a consensus risk window. Regulatory: US Section 232 (from 2026-01-15, 25% tariffs on certain advanced computing chips and derivatives; PCB bodies largely off the list, downstream finished goods restricted), continued US controls on AI chips to China, and rising US policy scrutiny of Chinese PCBs' role in the AI supply chain — if this tail risk materializes it would hit valuation rather than current orders (direct US export revenue share no more than 3%); tightening domestic environmental regulation forces out small players and benefits compliant top-tier capacity. The Thailand base (90%+ of overseas customers certified, already profitable) provides tariff avoidance and supply-chain resilience.
Company's industry position (dual basis, side by side): Per CIC (Zhuoshi) Consulting (cited in the company's H-share prospectus, revenue for the 18 months ended 2025-06-30), the company is global No.1 in data center PCBs (10.3% share) and global No.1 in 22+ layer PCBs; per Prismark's total-revenue ranking, the company's 2025 revenue of ~USD 2.6–2.7bn is at the edge of the global first tier (~3.1% of global output), below Avary (nine consecutive years at No.1, ~7%), Dongshan, and Shennan. The two bases are not contradictory: WUS is not No.1 in total revenue, but is global No.1 in the "data center high-multilayer high-speed board" segment where it actually competes, with share trending up (2026H1 datacom +73.5%, faster than the segment's +43.6%). Together with Victory Giant it forms the "AI PCB duo": Victory Giant is stronger in AI server compute cards/UBB with steeper elasticity; WUS tops switch/router high-speed boards with a more diversified customer mix and higher gross margin (38.7% vs 33.2%). Moat sources: certification barriers + process yield + scarcity of high-end capacity + focus strategy (no commoditized low-end expansion).
Quantified chain of the demand inflection: Driver trajectory (CSP 2026 capex USD 602–725bn → 2027E USD 1.2–1.3tn; Nvidia data center FY2027Q1 revenue USD 75.2bn, +92%) × unit content change (AI server accelerator boards 18–22 layers M8 vs general-purpose 8–12 layers; GB200 PCB value per rack ~RMB 147k → Rubin ~RMB 410k, sell-side estimates +179% to +233%; 1.6T switch PCBs 16–20 layers M9 vs 800G 12–14 layers) × penetration ramp (800G ports over 25% of total in 2025; 1.6T ~5–10% in 2026 → rapid ramp 2027–2028; AI server penetration lifted by +41% growth in global new data center capacity in 2026) × company share (data center boards 10.3%, global No.1; 800G yield over 90%; 1.6T shipping in volume; RMB 17.6bn expansion to capture it). Each link's data comes from independent sources (Prismark/Dell'Oro/Nvidia filings/CIC/company announcements); weak links: the company's precise share in 800G/1.6T switch PCBs is undisclosed, and the 1.6T penetration figure is inferred.
A generic framework tends to treat PCBs as ordinary manufacturing, but the industry's three decisive variables deserve separate scrutiny:
First, technology-generation positioning matters more than capacity scale. Pricing power in the PCB industry comes from "who is first to build the next-generation product": yield at 22+ layers, processing of M8/M9-grade materials, and mSAP fine-line capability determine a plant's tier in the AI board supply chain. Evidence of WUS's positioning: 800G switch board yield above 90%, capability up to 38–46 layers, 1.6T shipping in volume, 224G in small batches, 448G/3.2T in pre-research — a product in hand for every leg of the generational relay. This is the fundamental reason WUS's gross margin (38.7%) is markedly higher than the larger Avary's (21.5%) in the peer table: scale does not equal bargaining power; generation does.
Second, utilization and order visibility are the amplifiers of near-term earnings. In 2026Q1 the company's utilization was ~99%; with "high-end capacity tight," capacity is prioritized for top customers — at full load, incremental orders convert directly into profit (operating leverage). Tracking handles: the quarterly pace of construction-in-progress transfers to fixed assets, Thailand expansion release, the Qing-Song pilot production timing (2026H2), and shifts in "utilization/strong order book" wording in IR notes.
Third, cost pass-through determines the direction of gross margin. Raw materials (mainly CCL) account for over 50% of COGS, with upstream prices up cumulatively ~+50% over eight rounds within the year: pass-through capacity = pricing power × time lag. The contrast in samples is stark: datacom boards passed through successfully (gross margin +4.60pp), automotive boards failed (-6.22pp) — two outcomes within the same cycle, the difference lying in downstream supply-demand. From July 2026 the company raised prices 20–30% for telecom/automotive customers (per industry reports); if most of the increase is retained in 2027 contract renewals, the base-case gross margin of 36–37% is supported; if upstream keeps raising while downstream acceptance weakens, the automotive-board script takes over. Tracking handles: frequency of Kingboard/Shengyi price-increase letters, the company's sequential quarterly gross margin, and automotive board gross margin (leading sample).
Overall assessment: Cautiously bullish, confidence 0.55, time horizon 12 months. Core logic chain: earnings acceleration has been verified (Q3 preannouncement +100%+) → consensus lags by 10–15%, leaving room for upward revision → the current price of 18.7x 2027E embeds a conservative assumption that 2027 marks the earnings peak → demand-side evidence for 2027 (capex consensus +30%+, 1.6T volume ramp starting point) does not support that assumption → base-case fair value of RMB 114–140. Reasons for not setting confidence higher: high sensitivity to exit multiple assumptions (fair value of RMB 98–115 under a 12–16x framework), one-sided expectations structure (39 buy / 0 sell) and the market has already responded with -6.9% after the preannouncement, and the H-share issuance is in progress. Strategic implication: the current position is suited for accumulating in batches on pullbacks rather than chasing highs; the October 28 Q3 report and the CSP capex guidance for January–February 2027 are the two decisive validation points; if the stock first pulls back below RMB 95 (bear-case zone migration signal), assumptions must be re-examined; if it rises to RMB 140 (upper bound of base-case fair value), the margin of safety is exhausted and a revaluation is needed.
Risk and catalyst summary (written): Downside — ① industry capacity expansion wave in 2027–2028 plus decelerating capex triggering a price war (bear case RMB 55–82); ② upstream CCL price increases continue while pass-through fails (every -1pp in gross margin ≈ net profit -RMB 260 million); ③ widening FX losses (already RMB 276 million in H1); ④ H-share dilution of 5%–11% and potential discount; ⑤ inventory of RMB 6.04 billion facing concentrated impairment write-downs when the cycle turns down; ⑥ caveats attached to the RMB 2.2 billion dividend proposal (major-investment preconditions in tension with the RMB 17.6 billion capacity expansion). Upside — ① the October 28 Q3 report lands at the upper end of the preannouncement with gross margin holding at 38%; ② H-share hearing/pricing concluded, introducing an international anchor; ③ Thailand's quarterly profit expands sequentially (Q2 base of RMB 43 million); ④ Qingsong project trial production on schedule in 2026H2; ⑤ January–February 2027 CSP capex guidance exceeding US$1.3 trillion; ⑥ dividend proposal implemented via shareholders' meeting.
This report is based on public information and company announcements; the frozen price anchor is the closing price of 2026-10-23. Forward-looking estimates in this report (earnings forecasts, scenario fair values) are the analytical framework of this report and not company guidance; the sell-side consensus snapshot is as of 2026-10-23.