Rating: Neutral | Target price range: RMB 201–241 (base-case fair value, 12 months) | Current price: RMB 232.98 (close on 2026-09-01) | Total market cap: approx. RMB 229.0 billion (total A+H share capital of 983 million shares) | Margin of safety: -13.7% (base-case fair value floor of RMB 201 vs. current price) | Probability-weighted fair value approx. RMB 234, essentially in line with the current price
Victory Giant Technology is a global top-tier supplier in AI compute PCBs, a structurally expanding track: NVIDIA's Rubin platform ramping in 2026H2 brings a generational leap in PCB value per rack, and the company's order backlog extends through 2027, making it the highest-quality name in the A-share PCB sector (2025 ROE of 33.8%, gross margin of 35.2%, and No. 1 globally in AI compute PCB revenue). However, growth downshifted in 2026H1 — revenue +28.8%, net profit attributable to shareholders +33.3%, but non-GAAP (recurring) net profit only +12.5%, and Q2 recurring net profit -5.28% YoY — management attributes this to upfront headcount buildout and depreciation from new plants, with profit recovery conditioned on "revenue stepping up to the next level." The current price of RMB 232.98 implies 45.6x PE(TTM) and 27.3x on this report's 2027E earnings, which back-solves to an implied 2027 attributable net profit of approx. RMB 8.8 billion: the market has already marked down roughly 40% ahead of sell-side consensus, but still requires growth to reaccelerate. Base-case fair value is RMB 201–241, with the current price in the upper part of the range — no margin of safety; probability-weighted fair value of approx. RMB 234 is essentially at the current price, odds are nearly symmetric, and the bet is effectively a conviction play on "Q3 gross margin inflection + Rubin volume pass-through." Conclusion: high quality, fair-to-full valuation; maintain Neutral and reassess after Q3 results in late October validate the thesis or the price pulls back.
Nature of opportunity: A conditional-trigger opportunity, not an opportunity at the current price. The AI PCB volume-and-price logic is real (NVIDIA has officially confirmed Rubin is in full production, with products launching in 2026H2; the company's backlog extends to 2027), but the current price is about 5% above the base-case fair value midpoint of RMB 222. The entry hurdle is the Q3 2026 report in late October validating the gross margin inflection and the return to positive recurring profit growth, combined with Q4 Rubin revenue pass-through materializing.
Evidence grade: B. Key demand-side and share figures partly rely on third-party estimates (per-rack value is a Morgan Stanley supply-chain teardown estimate; "No. 1 globally" comes from an F&S commissioned category in the prospectus), and there is a measurement gap between company disclosures (top-5 customers at 41.98%, largest at 14.97%) and broad media estimates (NVIDIA chain exposure above 70%). This report primarily relies on disclosed figures.
Most likely material changes in the next 6–12 months:
Primary share price drivers (by importance): ① Revenue pass-through from Rubin ramp (determines whether 2027E lands at RMB 7.6 billion or 9.4 billion); ② Gross margin inflection (Q3 vs. Q2's 32.1%); ③ Share at major customers and next-gen platform cadence (orthogonal backplane positioning vs. rumors of Kyber delay as a two-way variable).
Expectation gap: Status — "the disagreement has mostly been absorbed by price." The current price is about 5% above the base-case fair value midpoint and essentially at the probability-weighted fair value; "a 40% discount to sell-side consensus" is not a margin of safety but the market's pricing of H1's realization/conversion rate (revenue +28.8%, Q2 recurring profit -5.28%). The remaining odds depend on two directional validations in Q4; at present there is no configurable odds advantage.
The AI platform generational transition is repricing PCB demand: per Morgan Stanley's supply-chain teardown (relayed by Cailianshe and EET-China, not first-party NVIDIA or company disclosure), PCB value per Rubin VR200 rack is approx. USD 116,700, a 233% increase over the approx. USD 35,100 for GB300 racks, adding two entirely new board types — midplane and ConnectX module boards; NVIDIA has officially confirmed Rubin is in full production, with products launching in 2026H2. Industry figures: the global PCB market at USD 85.836 billion in 2025 (+16.7%), 2026F approx. USD 101.95 billion (+18.8%) (Prismark); traceable figures put AI server PCB 2024–2029 CAGR at approx. 18.7% (Prismark, relayed), while the higher 2025–2030 CAGR of 24.8% is relayed data pending verification. Evidence of Victory Giant's capture: at the 2026-08-27 earnings call, management stated "next-generation products have entered volume production" and "high layer-count, high-end HDI orders have been scheduled through 2027" (management statement, not audited backlog).
Two discounts that must be stated honestly: first, the per-rack value leap is an industry-level variable; Victory Giant's share in VR200's new board types is unconfirmed by the company, with management saying only "share with leading overseas key customers has remained stable." Second, "approx. 50% share in GB200/GB300" and "NVIDIA orders exceeding 70% of revenue" are media or institutional estimates, not comparable with annual-report disclosures (top-5 at 41.98%). The confidence in this thesis has been haircut for source hierarchy.
2026H1 revenue was RMB 11.629 billion (+28.77%) and attributable net profit RMB 2.857 billion (+33.30%), but recurring net profit of RMB 2.418 billion grew only +12.50%, including Q2 recurring profit of RMB 1.161 billion, -5.28% YoY (vs. RMB 1.225 billion a year earlier); H1 gross margin 33.22% (-3.00pp YoY), Q2 32.10% (-6.72pp YoY). Management's official explanation for the Q2 pressure is two-fold: headcount grew over 30% in H1 (many employees still in training/reserve), and depreciation increased as new plants and equipment were progressively placed in service — "labor and capacity must stay ahead of orders." Copper-clad laminate price increases with lagged pass-through were listed as a risk in the interim report ("pass-through downstream involves a certain lag"), though management also said "raw material price volatility has a limited impact on gross margin" and "cooperation with key customers maintains a relatively stable gross margin level." Another layer of the haircut: H1 attributable profit included RMB 439 million of non-recurring gains (15.4% of attributable profit), mainly a RMB 595 million fair-value gain on allotted shares held (18.06% of total profit, which the company itself assessed in the interim report as "not sustainable"). Validation point: if Q3 gross margin rebounds above 33.5% and quarterly recurring profit turns positive YoY, the "temporary downshift" thesis holds; if it continues declining, the peak-profit narrative strengthens.
This report's 2027E attributable profit range is RMB 7.6–9.4 billion, midpoint RMB 8.4 billion, extrapolated from the H1 run rate plus capacity ramp: Q2 attributable profit of RMB 1.568 billion annualizes to approx. RMB 6.27 billion as the 2026 base, with 2027 increments from full-year Rubin volume ramp (revenue pass-through lagging rack shipments by about one quarter), the Thailand A2 site ramping, and the domestic compute/ASIC second curve. The engine mix must be presented honestly: management explicitly stated "major customers are prioritizing domestic production for orders," Thailand is taking only "orders from specific customers whose end products face restrictions," and ASIC/domestic compute "is growing faster but is still small in scale" — i.e., the RMB 8.4 billion burden rests mainly on the single engine of Rubin pass-through. The company's equity incentive assessment (announced 2026-07-20) trigger tiers require net profit or revenue growth of ≥40%/96%/174.4% in 2026/2027/2028 (translating to net profit trigger floors of approx. RMB 6.04/8.45/11.83 billion off the 2025 attributable base of RMB 4.312 billion), but the assessment is "net profit or revenue, whichever is met," the base excludes share-based payment expenses, and the grant price of RMB 155.94 is at a 30%+ discount to market — the market generally views it as a conservatively set exercise hurdle rather than profit guidance, so it serves only as a soft reference. Countervailing evidence is sell-side consensus (2026E RMB 8.68 billion, 2027E RMB 14.81 billion, n=10, already cut 12.9% within 90 days): hitting 2026E requires H2 attributable profit of RMB 5.82 billion, +104% HoH, clearly inconsistent with the Q2 momentum of negative recurring profit growth — consensus will most likely keep being cut down toward this report's range.
The current price of RMB 232.98 (close on 2026-09-01) corresponds to a market cap of RMB 228.97 billion, 45.6x PE(TTM) (TTM attributable profit RMB 5.026 billion), and 6.04x PB (book value per share RMB 38.57); on this report's earnings, 36.9x 2026E (RMB 6.2 billion) and 27.3x 2027E (RMB 8.4 billion). Back-solving implied expectations: at the base-case exit multiple of 26x, the current price implies 2027E attributable profit of approx. RMB 8.8 billion — about 4% above the incentive trigger floor of RMB 8.45 billion, about 46% above the 2026 trigger floor of RMB 6.04 billion, and about 41% below sell-side consensus of RMB 14.81 billion. Market pricing has already cut ahead of the sell side (the stock is down 40.7% from its 52-week high of RMB 392.95, consensus down 12.9% over 90 days, J.P. Morgan cut its H-share target from HKD 600 to 500 on 7-21 with earnings estimates lowered 5–13%, Nomura cut from HKD 479 to 447 on 7-15, lowering estimates 7–14%), but still requires growth to reaccelerate from H1's +33% while net margin holds at 24%. Zero-growth earnings power value (EPV) is approx. RMB 71/share (normalized EPS RMB 6.0, WACC 10%, net cash per share RMB 11.2), with the growth option accounting for approx. 69% of the current price — this ratio is sensitive to base selection (normalizing on consensus 2026E EPS of RMB 8.84 gives EPV of approx. RMB 105 and an option share of 55%), and given the company is at a structural inflection, a high option share is not by itself a cheap-or-expensive verdict; that is determined by the scenario distribution. Three scenarios: bear RMB 105–145 (2027E RMB 6.0–7.0 billion × 17–20x, -47.6% to -37.8% vs. current), base RMB 201–241 (2027E RMB 7.6–9.4 billion × 26x, -13.7% to +3.4%), bull RMB 290–395 (2027E RMB 11.0–13.0 billion, i.e., 74–88% realization of street consensus of RMB 14.8 billion × 26–30x, +24.5% to +69.5%); probabilities 22%/50%/28%, weighted fair value approx. RMB 234, +0.4% vs. current. The 26x exit multiple anchor is independent of the company's current price: it reflects a substantial discount to both the J.P. Morgan H-share framework (24x average FY26/27 EPS) and peers' current TTM multiples (WUS 45.9, Shennan 57.9, Shengyi Electronics 50.0, Avary 54.0) — peer multiples themselves embed continued AI-cycle expectations, and if earnings are cut, peer multiples would compress in tandem; this anchor's limitation is noted honestly.
First, customer concentration doubled in one year: the top-5 customer share rose from 20.03% in FY2024 to 41.98% in FY2025, and the largest customer from 5.73% to 14.97% (RMB 2.887 billion); combined with AI products accounting for about half of revenue (management statement), dependence on leading AI platforms constitutes single-customer risk — the 2026-06-23 rumor that "NVIDIA demanded a 10% price cut / share reallocation" once triggered a one-day -7.7%, showing the market's high sensitivity to such information; second-source diversion amid peers' capacity expansion (WUS separately announced RMB 17.7 billion of expansion, plus Shennan and Shengyi Electronics) is a persistent threat. Second, the supply cycle: 2025–26 global capex by leading PCB makers totals approx. USD 21.06 billion (industry media aggregation; Victory Giant's approx. USD 3.63 billion ranks first), plus the company's own 2026 investment plan of up to RMB 18 billion and 2026H1 free cash flow of -RMB 4.165 billion — annualizing H1 operating cash flow (approx. RMB 5.8 billion) against RMB 18 billion capex implies full-year FCF of approx. -RMB 12.2 billion, and net cash (approx. RMB 11.0 billion as of 2026-06-30) could fall to zero by end-2026; continued expansion in 2027 would depend on operating cash flow delivery or refinancing. If the AI capex growth downshift coincides with concentrated capacity release, volume and price would both come under pressure. A timing caveat: high-end laminate/board tightness is expected to persist through 2026–2027 (due to expansion delivery and qualification cycles), so the risk window opens after capacity is concentrated and placed in service in 2027–28. Governance watch point: on 2026-08-28 the controller, Chen Tao, transferred most of the equity in two holding platforms to his spouse (the company calls it a family arrangement with an acting-in-concert agreement signed and control unchanged); any subsequent equity changes will be amplified by market interpretation.
| Metric | FY2023 | FY2024 | FY2025 | 2026H1 | 2026Q2 (single quarter) |
|---|---|---|---|---|---|
| Operating revenue (RMB 100mn) | 79.31 | 107.31 | 192.92 | 116.29 | 61.10 |
| Revenue YoY | — | +35.31% | +79.77% | +28.77% | +29.49% |
| Net profit attributable to parent (RMB 100mn) | 6.71 | 11.54 | 43.12 | 28.57 | 15.68 |
| Attributable net profit YoY | — | +71.96% | +273.52% | +33.30% | +28.29% |
| Non-GAAP net profit (RMB 100mn) | — | 11.41 | 43.04 | 24.18 | 11.61 |
| Non-GAAP net profit YoY | — | — | +277.07% | +12.50% | -5.28% |
| Gross margin | 20.72% | 22.72% | 35.22% | 33.22% | 32.10% |
| Net margin (attributable/revenue) | 8.46% | 10.76% | 22.35% | 24.57% | 25.67% |
| Operating cash flow (RMB 100mn) | 12.80 | 13.58 | 46.03 | 29.11 | — |
| Free cash flow (RMB 100mn, OCF – purchases of long-term assets) | — | 5.25 | -20.14 | -41.65 | — |
| Cash + cash-like assets (RMB 100mn, period-end) | — | — | — | 214.12 | — |
| Interest-bearing debt (RMB 100mn, period-end) | — | — | 64.46 | 103.71 | — |
| Debt-to-asset ratio (period-end) | 56.13% | 53.44% | 52.85% | 39.85% | — |
Note: 2026Q2 single-quarter figures are derived as H1 minus Q1; balance-sheet stock items are period-end values for each period (2026H1 = 2026-06-30). Cash-like assets = cash of RMB 6.072bn + trading financial assets of RMB 15.339bn (incl. ~RMB 15.1bn of wealth management products and ~RMB 1.4bn of placement shares); restricted cash was only RMB 72mn.
Drivers of key metric changes (items with YoY ≥ ±20%): ① FY2025 revenue +79.77%: strong AI computing demand + ramp of high-end products; volume fell -2.72% (8.66mn sqm) while revenue surged—estimated ASP rose from ~RMB 1,129/sqm to ~RMB 2,087/sqm (+85%)—lower volume, higher price, with mix shifting toward high-end AI boards; ② FY2025 attributable net profit +273.52% / gross margin +12.5pp: the company attributes this to a significantly higher share of high-end products and mass production of 6-order 24-layer HDI; ③ 2026H1 finance costs +694.35% (RMB 297mn): the company explains this as FX losses and higher interest expense; ④ 2026H1 inventory +86.24% from the start of the year (RMB 5.89bn): management states this is proactive stocking in preparation for large-scale mass production of new products in H2; ⑤ 2026H1 operating cash flow +144.61%: increased sales collections; ⑥ 2026H1 net investing cash outflow of RMB 21.959bn / net financing inflow of RMB 22.147bn: higher purchases of long-term assets and wealth management products + receipt of H-share offering proceeds (net ~HKD 19.889bn).
2026H1 revenue was RMB 11.629bn (+28.77%), a marked deceleration from full-year 2025 (+79.77%); attributable net profit of RMB 2.857bn (+33.30%) included RMB 439mn of non-recurring gains—mainly RMB 595mn in fair-value gains on held placement shares (the company itself deems this unsustainable)—ex-out of which non-GAAP growth was only +12.50%. Q2 single quarter is the key deterioration signal: revenue of RMB 6.110bn was +10.7% QoQ and +29.5% YoY, but non-GAAP profit of RMB 1.161bn fell -5.28% YoY (revenue up nearly 30% while non-GAAP profit declined), with gross margin at 32.10% (-6.72pp YoY, -2.36pp QoQ). Management attributes this to headcount buildup (staff up over 30% in H1, many still in training) and front-loaded depreciation from new plants and equipment coming into service—"labor and capacity are running ahead of orders"—and promises "gross margin should move up in step as revenue steps up." This is a conditional proposition: if H2 Rubin stocking and delivery (conversion of the RMB 1.98bn half-year inventory build) fails to push revenue up a step, the depreciation and labor absorption logic does not hold. Against market expectations: the sell-side consensus of RMB 8.68bn for 2026E implies H2 attributable profit of RMB 5.82bn (+104% QoQ), while the company's incentive-trigger floor implies ~RMB 3.18bn for H2 (+11% QoQ)—a gap of over 80% between the two anchors is precisely what the Q3 report in late October will adjudicate. There is no reliable earnings-preannouncement baseline to judge "beat or miss"; this report only states: Q2 data ran below the trajectory implied by consensus at the start of the year, and the market has responded with downward revisions and multiple compression (YTD -24.85%).
Business model: asset-heavy electronics manufacturing (PCB), order-driven, non-subscription/non-recurring revenue; products center on high-layer-count boards (HLC) and high-order HDI, with AI accelerator cards/UBB/switch boards as the core incremental driver; exports account for 76.8%. Pricing power signals are strong—FY2025 estimated ASP +85%, and management says "every product iteration comes with a multiple increase in unit price"; but it is at times passively exposed upstream to high-end copper-clad laminate (Kingboard raised prices nine times from Aug 2025 to Jul 2026; Shengyi and Jinan Guoji raised Q3 prices 10-15%), and maintaining "relatively stable gross margins" with major customers limits pass-through headroom.
Cash content of earnings: OCF/attributable profit over four consecutive periods: 1.91 (FY2023) → 1.18 (FY2024) → 1.07 (FY2025) → 1.02 (2026H1)—always above 1, indicating good collection quality, but the ratio has declined steadily and the company has not explained the deterioration in cash conversion. FCF/attributable profit went from 0.45 in FY2024 to -0.47 in FY2025 and -1.46 in 2026H1—book profits are currently fully swallowed by capex. This is characteristic of an expansion phase rather than a red flag for fraud, but it means shareholders get no free cash flow for nearly two years. Decomposition of recurring earnings: of the 2026H1 attributable profit of RMB 2.857bn, the RMB 439mn (15.4%) gap versus non-GAAP profit of RMB 2.418bn comes mainly from fair-value gains on placement shares (one-off); core earnings growth should be judged on the non-GAAP basis (+12.5%), significantly below the headline attributable growth (+33.3%).
Return on capital: FY2025 ROIC ~23.7% (average invested capital, beginning and end of period), ROE 33.8%—far above WACC (10% in this report)—maintaining excess returns amid intensive capacity expansion is a genuine moat signal; but note the invested capital base is being rapidly inflated by RMB 18bn of 2026 capex, so from 2027 the ROIC denominator jumps sharply while the numerator depends on ramp speed.
Maintenance capex interrogation: CapEx/depreciation jumped from 1.04 in FY2024 to 6.52 in FY2025 (construction in progress rose from RMB 260mn to RMB 3.61bn; RMB 5.22bn in progress pending transfer at end-2026H1)—the current capital-black-hole configuration is clear: each year's earnings must be reinvested into capacity. The key differentiator from inefficient maintenance spending is order visibility (booked through 2027) and generational ASP leaps; if delivered, current capex is offensive rather than maintenance. If demand weakens, front-loaded depreciation will become a profit headwind (2027 depreciation could jump to RMB 2.5-3.0bn/year, per market estimates).
Red flags: ① customer concentration doubled in one year (top five 41.98%, largest 14.97%); ② persistently negative FCF, potentially reaching ~-RMB 12.2bn for full-year 2026 (net cash could go to zero); ③ nearly one-fifth of 2026H1 profit came from mark-ups on held shares; ④ inventory +86% in six months, with impairment exposure if demand weakens (H1 provision of only RMB 14.85mn already taken).
Credibility: pragmatic, with a preference for conservative targets. The 2024 annual report's operating plan aimed to "strive to reach the second RMB 10bn revenue milestone in 2026"—the company hit RMB 19.292bn in 2025, a year early; the RMB 100bn output target for 2030 cannot be verified. However, the equity incentive assessment thresholds (2026 trigger tier +40%) are significantly below institutional expectations (~+76%~108%), and the company refused to raise them, citing "comprehensive consideration of industry cycles and capacity deployment pacing"—an excellent track record and a conservative guidance habit, a "low-commitment, high-delivery" type—but it also means the incentive thresholds should not be read as earnings guidance.
Shareholder friendliness: neutral to dilutive. Payout ratios: FY2023 24.3%, FY2024 22.3%, FY2025 40.4% (RMB 20 per 10 shares, RMB 1.74bn total; no dividend in 2026H1); no cancellation-style buybacks (only RMB 41.68mn repurchased for incentive purposes). Frequent financing: 2015 IPO RMB 580mn, 2017 private placement RMB 1.08bn, 2021 placement RMB 2bn, 2025 placement RMB 1.9bn (at RMB 248.02/share), April 2026 H-share offering of 110mn shares (HKD 209.88, net proceeds ~HKD 19.889bn; H shares now 11.2% of total share capital)—three large equity raises within five years diluted existing shareholders, but bought capacity and order visibility, partly offset by the 18pp jump in the 2025 payout ratio. The ESOP (up to RMB 700mn, purchased in the secondary market, 12-month lock-up) is a confidence signal during the share-price drawdown.
Risk signals: on key questions—NVIDIA orders, share, Kyber delays—the company uniformly deflects with "commercial policy prevents discussing specific customers"; information transparency is poor. On 2026-08-28 the controlling shareholder transferred interests in a holding platform to a spouse (family arrangement + concert-party agreement); control is unchanged, but this leaves a governance observation point. Related-party transactions are minimal (only RMB 3.49mn of liquor purchases); key-management compensation totaled RMB 21.76mn.
| Segment | Revenue share (FY2025) | Gross margin (est.) | YoY | Business logic in one line |
|---|---|---|---|---|
| PCB manufacturing | 93.74% | ~32.0% | +79.92% | High-layer-count + HDI; global leader in AI accelerator/UBB/switch boards; among the first worldwide to mass-produce 6-order 24-layer HDI |
| Others (materials trading, etc.) | 6.26% | 84.7% (trading revenue-minus-cost basis) | +77.55% | Low-share ancillary trading; not a profit driver |
The company has no reportable segments (explicitly noted in the annual report), with a single core business: profits come almost entirely from PCB manufacturing (the other segment contributes only ~2% of gross profit). The annual report does not split revenue between multilayer boards and HDI; AI products are roughly half of revenue (per IR), while non-AI businesses (FPC/auto etc.) are stable in absolute terms—meaning all growth and gross-margin elasticity come from the mix shift toward high-end AI boards (the source of FY2025 ASP +85%). The stark gross-margin gap between segments (84.7% vs 32.0%) is an artifact of accounting bases (different revenue recognition for trading) and does not indicate relative business quality; the structural variables that truly need tracking are the AI share and high-order HDI share within the PCB segment, which determine whether the 32% gross margin is a floor or a midpoint.
Accounting red flags: ① profits include large fair-value gains on stocks (severity: medium)—trading financial assets included stocks of RMB 1.398bn at period-end, with +RMB 678mn in current-period fair-value changes booked to P&L; a further RMB 2.683bn stake in Founder Technology is booked in OCI; in 2026H1 fair-value gains of RMB 595mn were 18.06% of total pre-tax profit, and the company itself deems this "unsustainable"—a share-price decline in the holdings would claw back profits. ② Inventory growth far outpacing revenue (severity: low)—inventory +86.24% vs revenue +28.77%; the company explains it as stocking for H2 new products (raw materials and WIP up), which is reasonable but amplifies impairment exposure in a downturn; H1 saw only RMB 14.85mn of provisions.
Cross-period consistency: ① the jump and pullback in gross margin 20.7%→22.7%→35.2%→33.2% is directionally consistent with management's period-by-period explanations (higher high-end mix → front-loaded labor/depreciation); judged "consistent"; ② OCF/attributable profit declining from 1.91→1.18→1.07→1.02 with no company explanation of the deterioration in matching; judged "unexplained by the company"; ③ finance costs swung violently: FY2024 -60.35% → FY2025 +467.07% → 2026H1 +694.35%; the FY2024 annual report attributed the decline to "consolidation of acquisitions," when the main driver was actually FX swinging from a RMB 78.16mn gain to a loss (Note 66)—not fully consistent with subsequent explanations; judged "partially inconsistent."
Current market data and multiple percentiles: A-share close RMB 232.98 (2026-09-01, this report's price anchor), H-share HKD 215.0 the same day (~RMB 184.3); A/H premium of 26.4%—the pricing divergence itself is a valuation reference: the H-share market, where international institutions' target prices reside, is systematically below the A-share; if the spread converges, A-share upside compresses through the valuation channel even without earnings downgrades. Total market cap: RMB 228.97bn. PE(TTM) 45.6x (historical percentile: 2% over 1 year, 46% over 3 years, 68% over 5 years)—this percentile is for reference only, not a headline signal: attributable profit rose 6x in two years (RMB 671mn → 4.312bn), and the drastic change in the earnings base distorts the historical PE distribution; the 2% 1-year percentile is a double effect of "earnings explosion + share-price drawdown" (numerator and denominator) and cannot be read as cheap. PB 6.04x (50th percentile over 3 years) is relatively more reliable—the ROE trajectory (12% in 2022 → 33.8% in 2025) supports a PB above the 5-year median (4.27x). Forward multiples: this report's 2026E 36.9x / 2027E 27.3x; peer benchmarks (PE-TTM / 2025 revenue growth): Shengyi Technology (600183's peer WUS Printed Circuit) 45.9x/+42%, Shennan Circuits 57.9x/+32%, Shengyi Electronics 50.0x/+103%, Avary Holding 54.0x/+11.5%—Shenghong (Victory Giant) trades at the low end of the sector on multiples while having the highest earnings quality (ROE 33.8% vs WUS 28.6%, Avary 11.4%); a clear relative discount within the sector. The SW PCB industry median PE is 70.4x; the whole sector is in an AI re-rating regime, with absolute multiples generally not cheap.
Market-implied expectations: solving backward from a base-case exit multiple of 26x, the current price implies 2027E attributable profit of ~RMB 8.8bn—4% above the incentive-trigger floor (RMB 8.45bn) and 41% below sell-side consensus (RMB 14.81bn). In one sentence: the current price requires Rubin ramping on schedule, the Thailand ramp succeeding, and gross margin recovering—all three simultaneously; miss one and there is no margin of safety. Demand-side math can support high growth (2.8-3.3x jump in value per rack × NVIDIA data center revenue trajectory of +68% in FY2026 × order book extending into 2027), but the 2026H1 realized conversion rate (revenue +28.8%, gross margin -3pp) shows the value-per-unit leap did not translate linearly into profit—there is friction in pass-through (lagged price increases, front-loaded labor and depreciation, competition).
Three-layer value (EPV): asset value (book value per share) RMB 38.57, the liquidation floor; zero-growth EPV ~RMB 71/share (normalized EPS of RMB 6.0—based on 2026E recurring profit of ~RMB 5.9bn (close to the incentive trigger floor of RMB 6.04bn less interest income) ÷ 983mn shares, since a growth-phase company should not use historical averages; WACC 10%; net cash per share RMB 11.2 (2026-06-30 cash-like assets RMB 21.41bn – interest-bearing debt RMB 10.37bn)); the growth option (current price – EPV) is ~RMB 162, ~69% of the current price. Sensitivity noted honestly: if the normalized base uses consensus 2026E (EPS RMB 8.84), EPV rises to ~RMB 105 and the option share falls to 55%—"the share of price paid for growth" is sensitive to base assumptions and is not a standalone verdict on cheapness; cheapness is adjudicated by the scenario distribution.
Three scenarios and odds (12 months):
| Scenario | Probability | 2027E attributable profit | Fair range (RMB) | vs current | Key swing factor |
|---|---|---|---|---|---|
| Bear | 22% | RMB 6.0-7.0bn | 105-145 (midpoint 122) | -54.9% ~ -37.8% | AI capex digests in 2027 + further Rubin ramp delay + new industry capacity pressuring prices; gross margin falls below 30%; priced at 17-20x mean reversion |
| Base | 50% | RMB 7.6-9.4bn (midpoint 8.4bn) | 201-241 (midpoint 222) | -13.7% ~ +3.4% | Rubin ramps on schedule + Thailand starts up + gross margin recovers to 33-34%; 26x exit anchor: JPM 24x FY26/27 framework + substantial discount to peer TTM multiples |
| Bull | 28% | RMB 11.0-13.0bn | 290-395 (midpoint 342) | +24.5% ~ +69.5% | Street-delivery tier: 74-88% of consensus RMB 14.8bn + confirmed orthogonal backplane positioning; priced at 26-30x growth multiples |
Probability-weighted fair value is ~RMB 234, +0.4% vs the current price—the distribution is roughly symmetric around the current price (bear range -54.9% ~ -37.8%, bull range +24.5% ~ +69.5%); odds are neutral. The bear case vs the market's coordinates: the lowest publicly named major-bank target price is Nomura's HKD 447 (~RMB 383, still a Buy rating); 33/33 ratings are Buy or Overweight, none Neutral or Sell—this report's bear case (RMB 105-145) is significantly below the valuation implied by the most pessimistic published market forecast, and would require a substantive weakening of AI capex rather than sentiment swings; the base range (RMB 201-241) is also below the lowest target price (RMB 383)—rationale for going outside: the target-price framework was formed April-July and lags the August disclosures of Q2 non-GAAP turning negative and the H1 conversion-rate data; the current price (RMB 232.98, 52% of the median target of RMB 446.6) is itself the market's vote of no confidence in analyst targets; the bull case already incorporates the "street delivery" tier at 28% probability (RMB 11.0-13.0bn × 26-30x).
Our own earnings forecast (three-way comparison vs guidance and consensus):
| Basis | 2026E attributable profit | 2027E attributable profit | Notes |
|---|---|---|---|
| This report | RMB 6.2bn (5.81-6.51) | RMB 8.4bn (7.6-9.4) | H1 run-rate + ramp extrapolation; H2 attributable profit ~RMB 3.3bn (+17% QoQ) |
| Equity incentive trigger/target tiers | RMB 6.04 / 6.47bn | RMB 8.45 / 9.70bn | Either/or assessment + ex-share-based payment; soft reference |
| Sell-side consensus (n=10) | RMB 8.68bn | RMB 14.81bn | -12.9% over 90 days, still 40%/76% above this report |
Conclusion: fairly priced, no margin of safety. Assess quality (sector-best ROE/share/process generation gap) separately from price (growth already paid for): this is a company worth placing in a core watchlist, but buying at the current price is essentially a bet that both Q3-Q4 validation points deliver in full; expected return is near zero, with left-tail risk borne. Target range RMB 201-241 (base-case fair value); if the share price falls to the lower half of the range (below ~RMB 210) and Q3 confirms a gross-margin inflection, the odds turn favorable.
Industry Size: The global PCB market is expected to reach USD 85.836 billion in 2025 (+16.7%) and approximately USD 101.954 billion in 2026F (+18.8%), with a 2025–2030 CAGR of about 11%, reaching roughly USD 123.3 billion by 2030 (per the Prismark website and its 2026Q1 report); Mainland China accounts for 57.8% of the global market (USD 49.652 billion in 2025, CAGR 10.5%). The structural growth engine is AI: in 2024, the global PCB market grew only +5.8%, accelerating to +15.4%–16.7% in 2025, with server/data storage PCBs up +46.3%, while global electronic end products grew only +8.5% over the same period—virtually all industry growth was driven by AI servers and high-speed optical communications, with severe product-mix divergence (AI boards +46% vs. conventional FPC +3%). AI-server-specific PCB segment: approximately USD 16 billion in 2025, with a verifiable 2024–2029 CAGR of about 18.7% (Prismark, as relayed) and a higher estimate of a 2025–2030 CAGR of 24.8% (relayed, unverified); AI-related HDI has a CAGR of about 29.6%, and multi-layer boards of 18+ layers about 33.8%. Monthly data from the China PCB industry association (CPCA) and Taiwan's TPCA likewise point to high-end capacity utilization running at full load.
Quantified Chain of the Structural Demand Inflection: Driver = AI capex—NVIDIA data center revenue of approximately USD 115.2 billion in FY2025 → approximately USD 193.7 billion in FY2026 (+68%), and a single-quarter USD 89.0 billion in FY2027Q2 (+117%); the four major hyperscalers' 2026 capex guidance totals approximately USD 600–725 billion (vs. approximately USD 410 billion in 2025, +47%–77%); Unit consumption jump = PCB value per rack of approximately RMB 25,000 for GB200 → approximately RMB 36,000 for GB300 → approximately USD 116,700 for Rubin VR200 (Morgan Stanley supply-chain teardown: +233% vs. GB300, adding two entirely new board types—18 mid-planes at USD 1,500 each and 72 ConnectX module boards at USD 270 each; layer counts iterating toward 22+ layers / high-end 78 layers); Penetration pace = Rubin NVL144 entering mass production in 2026 (officially confirmed full production, launch in H2), followed by Rubin Ultra/Kyber in 2027, with 1.6T optical module PCB growth of approximately 700% in 2026 being the steepest curve; Company share = Victory Giant ranked No. 1 globally in AI and high-performance computing PCB revenue in 2025Q1 (F&S commissioned category), and No. 6–7 globally by total revenue (Prismark); media reports claim an approximately 50% share of GB200/GB300 server PCBs (unconfirmed by the company). Judgment on the nature of this chain: the unit-consumption-level jump (2.8–3.3x) combined with penetration ramp-up represents a structural rightward shift of the demand curve, not an ordinary restocking cycle—the company's revenue growth decelerating to +28.8% in 2026H1 precisely reflects conversion frictions (capacity, certification, price pass-through), not demand refutation; however, share and value distribution are the softest link in the chain (the 50% share is a media figure), so valuation sensitivity should be treated as a range.
Supply Chain and Value Distribution: Upstream copper-clad laminates (Kingboard No. 1 globally in rigid CCL, Shengyi Technology No. 2) and further upstream HVLP4 specialty copper foil, low-CTE glass cloth (dominated by Japanese players such as Nittobo), and high-speed resins—Kingboard issued nine rounds of price increase letters between August 2025 and July 2026 (the sixth round within 2026 raising FR-4/PP by another +15%, with only 20 days between rounds); Shengyi and Gold Coin (Jin'an Guoji) raised prices 10–15% in 2026Q3, and Mitsubishi Gas Chemical +30%—profits are flowing back upstream (Shengyi Technology 2025 net profit +92%, ROE 21%). Midstream PCB manufacturing (this company): AI high-end board makers have net margins exceeding 20% (Victory Giant 22.4%, WUS 20.2%), the highest in the manufacturing segment, but the nine consecutive CCL price hikes in 2026 face a lag in downstream pass-through, and the market is closely watching pricing pass-through capability. Downstream hyperscalers/ODMs (Foxconn, Quanta, Wistron) → NVIDIA/AMD/Broadcom/Google and other compute customers: certification cycles of 1–2 years and redone generation by generation; PCBs account for a low share of total BOM (GB200 PCB cost of approximately USD 407 per GPU vs. rack costs in the millions of dollars), so customers have low price sensitivity but high sensitivity to delivery certainty—midstream high-end board makers have strong bargaining power downstream but are periodically passive vis-à-vis upstream high-end materials, hedging through scale purchasing and long-term orders; Thai capacity still relies on core materials imported from the mainland, with costs 10–20% higher.
Supply-Demand and Competitive Landscape: On the demand side, as above (AI-driven structural shortage); on the supply side, this is the strongest capacity expansion wave in history—total industry capex of approximately USD 21.061 billion in 2025–26 (industry media aggregate, 82% concentrated in Mainland China + Taiwan); the top three expanders are Victory Giant at approximately USD 3.633 billion, WUS at approximately USD 3.009 billion, and ZDT at approximately USD 2.854 billion, all directed toward high multilayer HLC, mSAP HDI, and ABF substrates; but supply is constrained by four bottlenecks (HBM/memory crowding out, HVLP4 copper foil and low-CTE glass cloth shortages, advanced wafer restrictions, and lengthening equipment delivery lead times), and Prismark judges the high-end laminate shortage to persist through 2026–2027. Concentration: global PCB TOP40 revenue of USD 67.34 billion in 2025, with the top three being ZDT at USD 5.87 billion, Unimicron at USD 4.23 billion, and DSBJ at USD 3.57 billion; in AI high-end HDI/accelerator boards, the actual first tier comprises only a handful of players—Victory Giant, WUS, Unimicron, ZDT, etc.—with extremely high CR. Entry barriers: process-generation gaps (among the world's first to mass-produce 6-order 24-layer HDI, 100+ layer high multilayer, back-drilling/mSAP), NVIDIA ecosystem certification (1–2 years and redone generation by generation; Tier-1 positioning is hard to displace), capital thresholds (billion-level per plant), and materials collaboration (M6/M8 high-speed materials require joint development with CCL makers). Substitution threat: CPO co-packaged optics may in the medium-to-long term shift some switching/interconnect functions into the package (the company responded that it "will not fully replace backplanes, only used in specific scenarios"; current evidence actually shows the Kyber rack was delayed due to PCB mid-plane manufacturing bottlenecks); the ASIC route does not reduce PCB demand and instead broadens the customer base.
Cycles and Regulation: AI PCBs are in the mid-stage of a growth cycle (2023 introduction → 2025 explosion → 2026–27 Rubin volume ramp), while traditional PCBs are mature and oversupplied; the industry is in an upcycle but with decelerating marginal growth, and concentrated capacity release in 2027–28 is the biggest cyclical foreshadowing. Leading indicators: NVIDIA data center revenue and guidance, the four hyperscalers' capex, CCL price-hike frequency (interval shortened to 20 days), Rubin ramp pace, and equipment lead times. Regulation: under U.S. reciprocal tariffs, Thailand's rate is set at 19% (effective August 2025); USTR's Section 301 investigation into "structural overcapacity" across 16 Southeast Asian economies; U.S. Section 301 tariffs on Chinese semiconductors stacked up to 50% (proposed further increase from June 2027)—the company's direct U.S. export exposure is minimal (Mainland China's direct PCB exports to the U.S. were less than 1% in 2025Q1; trade terms stipulate tariffs borne by customers), so the net policy impact is limited, but policy uncertainty over Southeast Asian capacity and the 10–20% cost premium remain persistent friction items.
Peer Benchmarking: See the E6 benchmarking chart. By scale: Avary (RMB 39.15 billion) > Shennan (RMB 23.65 billion) > Victory Giant (RMB 19.29 billion) > WUS (RMB 18.95 billion) > Shengyi Electronics (RMB 9.49 billion, 2025 revenue); but the ranking by earnings quality and AI purity is almost reversed: Victory Giant posted the sector's highest 2025 ROE of 33.8% and net margin of 22.4%, while Avary (mainly Apple-chain FPC) net profit grew only +3.3% and Kinwong only +5.3%—"having AI high-end orders or not" has become the single dominant variable behind the sector's earnings divergence. Victory Giant and WUS are the twin AI PCB leaders: WUS excels in ultra-high-layer-count switch backplanes/networking boards with a more balanced customer mix, while Victory Giant leads in accelerator-card high-end HDI and ASIC elasticity; Shengyi Electronics benefits from Shengyi Technology's CCL group synergies with the steepest growth (+102.6%) but on a low base.
Company Positioning in the Industry: Global No. 1 in the AI compute PCB niche (F&S 2025Q1 basis; No. 6–7 globally by total revenue—the two rankings use different category widths and are presented side by side), with a rising share trend (revenue from RMB 10.7 billion in 2024 → RMB 19.29 billion in 2025 → +28.8% in 2026H1), and a triple moat: process-generation gap (first mass production of 6-order 24-layer HDI, 100+ layer capability, leading Rubin-generation R&D), NVIDIA Tier-1 certification stickiness (entered in 2020, certified generation by generation, orders scheduled through 2026), and capacity and globalization first-mover advantage (No. 1 global capex in 2025–26, Thailand A1 in mass production / A2 targeted for mass production in 2026Q3 / A3 in planning, Vietnam AI HDI project). Constraining factors: single-platform dependence, lag in upstream price pass-through, and the possibility of cyclical reversal from "shortage premium" back to "cost competition" after industry capacity release in 2027–28.
Overall Judgment: Neutral (confidence 0.60, 12 months). Victory Giant Technology is one of the highest-quality names in the A-share AI compute chain—a globally No. 1 niche position, the sector's highest ROE and net margin, orders booked through 2027, and a solid quantitative foundation for the structural demand inflection; however, the current price of RMB 232.98 (2026-09-01 close) has already fully priced in the base-case growth: PE(TTM) of 45.6x and 27.3x on 2027E, implying the upper end of the achievable range for approximately RMB 8.8 billion in 2027 net profit attributable to shareholders, with a margin of safety of -13.7%; the probability-weighted fair value of approximately RMB 234 essentially coincides with the current price. Strategically: do not chase highs, do not short—the downside scenario requires a substantive weakening of AI capex to trigger (current NVIDIA FY2027Q2 data center +117% does not support this), and the upside scenario requires a Q3 gross margin inflection and Q4 Rubin pass-through confirmation. Two re-evaluation triggers: if the stock falls below RMB 210 (lower half of the base-case range) and Q3 confirms a gross margin recovery, the odds improve; if the stock breaks below RMB 201 (base-case lower bound), the Rubin pass-through thesis must be re-examined for refutation.
Risk Disclosures (Written Summary): ① Customer concentration: top five customers account for 41.98% and rising; market sensitivity to NVIDIA-chain annual price cuts or second-source diversion rumors is extremely high; ② Gross margin: if CCL pass-through lag and front-loaded labor/depreciation are not digested in Q3, full-year net profit attributable could fall to RMB 7.5–8.0 billion (consensus downgraded another 10–15%); ③ 2027–28 supply release combined with deeply negative FCF (approximately -RMB 12.2 billion for full-year 2026; net cash may fall to zero); ④ H-share cornerstone lock-up of approximately 55–66 million shares expiring 2026-10-21, overlapping with the Q3 earnings window, with margin balance at 7.81% of free-float market cap amplifying volatility; ⑤ FX: a 5% CNY/USD fluctuation impacts net profit by approximately ±RMB 226 million (USD receivables of RMB 4.63 billion); ⑥ Technology roadmap: long-term CPO substitution for switch/backplane board volumes (to be monitored after 2027H2); ⑦ Governance watch: subsequent changes to the controlling family's equity structure adjustments.
Catalyst Calendar (Next 12 Months):
Core monitoring sequence: Q3 single-quarter gross margin and non-GAAP YoY (verification point one), Q4 revenue QoQ and share of high-multilayer/HDI revenue (Rubin pass-through, verification point two), top-five customer concentration trend, Thailand A2 mass production and overseas revenue share, frequency of CCL price increase letters and the company's price adjustments with major customers, and the pace of downward consensus revisions.
This report's price anchor is the 2026-09-01 A-share closing price of RMB 232.98; all earnings forecasts and scenario fair values are this report's own estimates. Data sources include the company's periodic reports, investor relations records, Prismark, and public reporting by named institutions.