2476.HK — Victory Giant Technology (002476.SZ)

Companion read: ai-server-pcb-primer for the full value chain, materials squeeze, and bottleneck framing. This page treats Victory Giant as a single name; the primer is where the supply-chain context lives. Voice: Register D.

Thesis

Verdict: BUY-QUALITY business, WATCH on entry. Tier 2 Tactical — not Tier 1 Core. Medium-high conviction on business quality, Medium on entry valuation. Victory Giant (胜宏科技) is the #1 global supplier of advanced PCBs for AI servers — 13.8% global AI/HPC PCB share in 1H 2025, up from 1.7% / #7 rank in 2024, a ~10x share gain in 12 months that is not normal in a mature industry. It is NVIDIA's named anchor PCB supplier for the GB200/GB300 generation and, per the 7 May 2026 CEO channel check, holds 50-55% of global Vera Rubin platform supply with volume shipments already begun Q2 2026 on 44-layer boards (Rubin GM >45%). It is the cleanest single-name way to own the AI accelerator capex cycle on the PCB layer.

The catch is concentration and price. HK prospectus disclosure confirms NVIDIA >60% of FY2025 revenue (>70% in Q1 2025); GB200/GB300 alone ≥50% of FY2025 revenue — materially higher than the original 40-55% estimate, and well above the "diversified" threshold. This is a single-customer-program-cycle bet dressed up as a public company; there is no version of the story where you get the upside without the concentration risk. The same concentration is what made the 10x share gain possible (focused R&D + capacity around NVIDIA's roadmap), so it is structural, not accidental.

The stock is rich post-rerate: ~US$40B market cap on ~US$1.5B FY2024 revenue, the A-share +580% in 2025 (led the MSCI Asia Pacific index), the HK secondary listing (21 Apr 2026, US$2.6B, largest HK IPO of 2026) closing day-one +57.23%. On a relative basis it is cheaper than Taiwanese/Japanese peers (Nanya 44x FY27 P/E, Unimicron 32x FY27 P/E) and the TAM tailwind is the largest in the sector (GS Jan 2026 revision: AI PCB TAM $26.6B in 2027, 140% CAGR). Relative valuation is the bull case; absolute valuation is the bear case. The Q1 2026 print (rev +28% YoY, NI +40% YoY, net margin 23.3% with ~30% incremental margin) strengthens the business-quality leg while doing nothing for the concentration leg. Own it at a price, sized below a single-name core, paired with the higher-leverage upstream bottleneck names (Fulltech, Kitagawa Seiki).

Conviction: Medium-high on business quality; Medium on entry valuation — post-rerate, narrow A/H spread, single-customer concentration hard to ignore. The HK listing brings Western capital onto the register but resets the stock as a momentum-trade candidate.

Snapshot

One-liner: the highest-AI-purity name in the global advanced-PCB stack — #1 in AI/HPC PCB share, NVIDIA's anchor supplier for GB200/GB300 and 50-55% of Vera Rubin platform supply, with a Huizhou + Thailand + Vietnam footprint that hedges US-China tariff risk; rich post-rerate and concentrated on a single customer.

• Tickers / exchange: 002476.SZ (Shenzhen primary, since 2015) / 2476.HK (Hong Kong secondary, 21 Apr 2026). GICS Technology / Electronics Hardware; industry = advanced PCB (anylayer HDI / HLC).
• Legal name: Victory Giant Technology (Huizhou) Co., Ltd. — 胜宏科技(惠州)股份有限公司; HQ Huizhou, Guangdong; founded 2003; FY-end Dec 31; shpcb.com
• HK price (latest): HK$319.4 | HK market cap HK$313.9B (~US$40B) | EV HK$282.6B (~US$36B) | HK-class shares 110.2M
• HK IPO: priced HK$209.88 / debut close HK$315 (+57.23%); ~37% discount to Shenzhen close (vs typical 15%); 431x retail oversubscription, 18.5x institutional book
• Stock perf: A-share 002476.SZ +580% in 2025 (led MSCI Asia Pacific Index; an earlier draft cited +600% — retained as the source-of-record figure superseded by +580%); HK debut +57%. 52-wk HK range n/a (just listed)
• Valuation: P/E (TTM) high (~80-100x on FY2024 NI; ~30-40x on FY2025E NI); EV/Revenue ~24x FY2024 / ~12-15x FY2025E; FCF yield negative (capex phase); dividend yield ~0.5-1% (A-share). HK-only multiples are partial — total A+H market cap is materially higher; treat as directional.
• FY2025 (verified, HK prospectus): revenue CN¥19.3B (+80% YoY); NI ~CN¥4.29B (+273%, per IT Business Net headline math)
• Conviction: Medium-high business quality / Medium entry valuation. Tier 2 Tactical.
• yfinance warning: ticker 002476.SZ resolves to an UNRELATED company ("Shandong Polymer Biochemicals") — treat yfinance SZ data as broken for this name. 2476.HK resolves cleanly. Use Wind / iFind / SZSE official disclosures / HKEX prospectus for A-share fundamentals.

Business

Victory Giant makes printed circuit boards — specifically the high-layer-count, anylayer-HDI, low-loss boards that route high-speed signals across NVIDIA AI accelerator boards, NVSwitch trays, hyperscaler custom silicon (Trainium, TPU), and the densest networking switches. About >70% of revenue is now AI/HPC — the highest mix in the peer set by a clear margin. Two things few Asian peers have managed: (1) climbed the layer-count ladder from 8-12L commodity boards (2010) to 30-40+L AI server boards (2025-26) in a 12-year run; (2) captured share from incumbents while doing it (#7 → #1, 1.7% → 13.8%).

Business / segment mix (single reported segment; decomposed from prospectus + Collyer Bridge / Acid Investments commentary, not officially segmented):

Business line	% of FY2024 rev (est.)	Notes
AI server / HPC PCB	~70-75%	NVIDIA HGX/B200/GB200 boards, NVSwitch tray, hyperscaler custom silicon
Networking switch / 5G base station PCB	~10-12%	Adjacent tech, similar materials, different customers
Automotive HDI PCB	~5-8%	Growing but small relative to AI
Consumer / industrial / other	~5-8%	Legacy, declining as % of mix

Asset-heavy fab (capex/revenue typically 12-18% in advanced PCB; VG running ~20%+ on parallel Thailand + Vietnam builds). VG's role is Layer 4 — PCB fabrication: it turns CCL into a finished, electrically tested board. It does NOT do component placement (Foxconn) or rack assembly. It owns the most technically demanding layer of the rack BOM after the silicon itself.

Product segments (deep-dive):

• A — AI accelerator OAM/UBB baseboards (HGX, GB200, GB300): ~600×400mm, 22-30 layers, anylayer HDI, Megtron 7/7N. VG primary or co-primary supplier. ASP $2,000-5,000/panel; 10-15M panels/yr industry-wide by 2026E.
• B — NVSwitch / network switch trays: the NVL72 switching fabric, 1.8TB/s NVLink, 30-50+ layers, M-grade. Per Collyer Bridge, VG gaining share against incumbent (rumored Tripod). Highest-layer-count, highest-ASP work in the rack BOM. ASP $5,000-15,000/tray; 100-200K trays/yr by 2026E.
• C — Vera Rubin midplane/backplane PCBs: per STF/GS, VR200/VR300 replaces bridge cables with PCBs/CCLs. Midplane content $171-256/GPU (107%/57% jump vs GB300); backplane content $781-1,563/GPU (5x/4.5x jump). Demand 2H26 (midplane) / 2H27 (backplane). VG position (updated 7 May 2026 CEO call): 50-55% of global Rubin platform supply, co-developed with NVIDIA from the start, volume shipments began Q2 2026 on 44-layer boards, Rubin GM >45%. This closes the prior "likely beneficiary, not yet confirmed" gap and reframes Segment C from speculative 2H27 upside to in-flight 2026 revenue — the single biggest 2026-27 catalyst, now de-risked on the supplier-status leg. Material: M9 + Q-Glass (narrowest processing windows, highest qualification barrier).
• D — Hyperscaler custom silicon (Trainium, TPU, Maia, MTIA): multi-customer, 18-24mo design cycles. The hedge against NVIDIA concentration. Google TPU specifically (7 May CEO call): high-end AI PCBs already in mass production; management guides several billion RMB FY2026 revenue from Google — no longer a roadmap line item. (Note: this claim is CEO-call-only — see research gaps.)
• E — Networking switch PCBs (800G, 1.6T): Arista/Cisco/white-box + NVIDIA Spectrum/Mellanox. Adjacent demand, stable.
• F — Automotive HDI PCB: small, cyclical diversification; not the thesis driver.

Operations footprint (updated 7 May 2026 CEO call):

Site	Role	Status	Notes
Huizhou, China (HQ)	Primary advanced PCB fab	Operating; Buildings 10 & 11 commissioning Jun-Aug 2026	Multi-plant campus. Bldgs 10+11 add combined RMB 40bn annual capacity (Bldg 10 equipment commissioning Jun-Jul, Bldg 11 Aug — 2-mo slip vs original plan)
Thailand (APCB / A1)	AI server PCB capacity, US tariff hedge	Operating; A1 producing high-end AI qualification boards	Acquired APCB's PCB ops 2024 (wholly-owned subsidiary, not a JV). Yields tracking toward Huizhou parity within 3-6 months; 2nd facility planned
Vietnam (Bac Ninh)	Greenfield AI server PCB plant for international customers	First building goes live H2 2026 (tightened from prior Q4 2026 estimate); greenfield $520M project, started March 2025	Capacity target RMB 11-13bn; serves non-China-origin customers

Thailand and Vietnam capacity is hard to read as anything other than a deliberate decoupling hedge against US 301 tariffs on China-origin PCB. No disclosed equity-level JVs with NVIDIA, Foxconn, or any hyperscaler — relationships are commercial supply agreements. Collyer Bridge's note that VG is referenced in NVIDIA's roadmap suggests design-in collaboration beyond arm's-length supply, but it is not a formalized equity partnership.

First principles — the technology (full treatment in ai-server-pcb-primer §III-VI): VG manufactures 20-40+ layer anylayer-HDI boards with stacked microvias, line/space down to 30/30µm, on premium low-loss laminates (Megtron 7/7N today, M9 + Q-Glass coming for Vera Rubin). Per the 7 May CEO call: 70-layer boards in volume production, 100-layer in reserve; management claims VG is the only vendor globally supplying 44-layer AI boards at scale; HDI 10-stage 30-layer in production, 14-stage 36-layer in pre-development — tracking NVIDIA's next-gen requirements 2-3 years in advance. The technical moat is yield at high layer count: industry yield falls ~73% (2025 mainstream) → ~62% (2027 Vera Rubin) per STF/GS; VG's position is contingent on holding yield 5-10pts above competitors at the same layer count, and the prospectus implies it does. Process know-how is NOT patentable — it is accumulated craft (clean-room discipline, lamination cycle control, drill setup, CMP recipes), which is exactly what makes a 12-year share-gain story defensible: it cannot be replicated by capital alone.

Customers & concentration (verified from HK prospectus):

#	Customer	Ticker	Revenue Share	Relationship
1	NVIDIA (direct + via Foxconn/Wiwynn)	NVDA	>70% in Q1 2025; >60% projected FY2025; GB200/GB300 alone ≥CN¥10B FY2025 (>50% of total)	Design-in / qualified vendor (HGX, B100, B200, GB200/GB300; Vera Rubin confirmed)
2	Foxconn (ODM on behalf of NVIDIA + others)	2317.TW	~10-15% (overlaps NVDA)	Build-to-spec, multi-year framework
3	Hyperscaler custom silicon	GOOGL / AMZN / MSFT / META	Google TPU alone: several B RMB FY2026 (mass production, per 7 May CEO call); AMZN/MSFT/META smaller	Project-by-project (Trainium, TPU, Maia, MTIA)
4	Networking OEMs	Cisco, Arista	small	Build-to-spec (800G / 1.6T)
5	Chinese-domestic AI	Huawei, Cambricon (unlisted/private)	small	Design-in (decoupling-era exposure)

Discrepancy flagged (kept both): the original 2026-05-03 deep-dive estimated NVIDIA at 40-55% of revenue and hyperscalers ~10-15% combined. The HK prospectus follow-up revised NVIDIA to >60% FY2025 / >70% Q1 2025 with GB200/GB300 alone ≥50% — materially higher concentration than first estimated. The verified prospectus figure governs; the 40-55% estimate is retained here only as the prior-draft anchor.

Concentration metrics (verified): Top-1 (NVIDIA) >60% FY2025 / >70% Q1 2025 (peer set Tripod/Unimicron typically 25-35%); NVIDIA GB200/GB300 product family alone ≥50% of FY2025 revenue; Top-3 ~80-85% (est.); Top-5 90%+ (est.). If NVIDIA walked tomorrow VG loses >60% of revenue and the highest-margin chunk. The mitigant is structural — NVIDIA cannot walk from a qualified, high-yielding fab in a tight supply environment without absorbing 12-24 months of qualification cost on alternates. The protection is not "won't" but "can't easily." Hyperscaler diversification (Google TPU now visibly multi-billion-RMB) is real but small relative to NVIDIA.

Financials

Multi-year (CN¥B; FY2025 verified per HK prospectus, rest estimated — yfinance SZ broken):

Metric	FY2022	FY2023	FY2024	FY2025 (verified)	FY2026E
Revenue	~5.5	~6.5	10.73	19.30 (+80%)	~33 (sell-side +70%)
Revenue growth YoY	n/a	~18%	65%	+80%	~+70%
Net income	~0.45	~0.55	1.15	~4.29*	~4.0-5.0
Net margin	~8%	~8.5%	10.7%	~22.2%	~13-15%
Total assets	~10	~14	19.18	~28+ post-HK	~40+
Total equity	~5	~7	8.93	post-HK ~25+	~30+

*FY2025 NI implied by "Net Profit Surged 273%" headline (1.15 × 3.73 = 4.29B per IT Business Net). An earlier draft estimated FY2025 NI at ~CN¥2.0-2.5B (net margin ~12-13%) before the prospectus/full-year print landed; the ~CN¥4.29B / ~22% verified figure supersedes it.

Core Four:

1. Organic revenue growth — spectacular: 65% FY2024, +80% FY2025, ~+70% FY2026E. The single most important metric for the thesis.
2. Margins — expanding and accelerating: net margin 10.7% (FY24) → ~22% (FY25) → 23.3% (Q1 2026), with incremental net margin Q1 2026 ~30% (above blended). AI server PCB carries higher GM than commodity; mix shift + operating leverage is the driver. Rubin GM >45% per CEO.
3. Capital intensity — high and rising, ~20%+ on parallel Thailand + Vietnam expansions.
4. Capital deployment — reinvestment-dominant; HK IPO proceeds = capacity expansion ammo; buybacks not in the plan.

Q1 2026 verified (SZSE filing, 29 Apr 2026):

Metric	Q1 2025	Q1 2026	YoY
Revenue	~CN¥4.31B	CN¥5.519B	+28.0%
Net income (attributable)	~CN¥0.92B	CN¥1.288B	+39.95%
Non-GAAP NI (扣非)	n/a	CN¥1.257B	+36.07%
EPS	CN¥1.07	CN¥1.48	+38.3%
Operating cash flow	~CN¥0.42B	CN¥2.117B	+399.4%
Net margin	~21.3%	23.3%	+200bps

Net-margin trajectory: FY2024 10.7% → FY2025 ~22.2% → Q1 2026 23.3% — margin expansion is accelerating into 2026 despite tougher comp. Incremental net margin Q1 2026 ~30% (ΔRev +CN¥1.21B / ΔNI +CN¥0.37B), suggesting marginal AI accelerator board mix is even more profitable than the FY25 average. ΔOCF +CN¥1.70B vs ΔRev +CN¥1.21B → operating cash conversion >100% of revenue growth = working capital improving, not burning cash on the ramp.

One-time distortions: Q1 2026 OCF surge (+399%) reflects favourable working-capital release (Rubin customer prepayments, lengthened payables), not steady-state. Thailand/Vietnam ramp opex sits below the line and won't show as P&L drag until 2H 2026. (Earlier draft flagged Vietnam ramp + APCB integration costs as likely depressing 2024-25 reported margins below true run-rate.)

Cash flow & balance sheet: OCF positive but capex absorbs most; FCF likely negative-to-barely-positive through the 2024-27 expansion phase. 2026 capex guide ≤RMB 20bn, funded entirely by operating cash flow (7 May CEO call) — HK IPO proceeds + OCF cover Bldg 10/11 + Thailand + Vietnam without further equity raises, eliminating the near-term dilution risk earlier flagged as "ATM/shelf unverified." Net debt low pre-IPO, near-zero post-IPO with the US$2.6B raise; net debt/EBITDA well under 1x. ROIC estimated 12-18% and rising — high for advanced PCB peers.

Implication for thesis: the Q1 2026 print strengthens the business-quality leg (margins expanding, OCF self-funding) while doing nothing for the concentration leg (NVIDIA still >60%). Conviction on quality nudges higher; entry-valuation stance unchanged.

Industry landscape

A pure-play on the AI server PCB inflection riding the global AI capex super-cycle. Full value-chain, materials-squeeze, and bottleneck framing live in the companion primer — see ai-server-pcb-primer. VG sits at Layer 4 (PCB fab); the biggest mispricings sit two layers upstream — glass cloth (Fulltech 1815.TW, mkt cap <$1B) and vacuum press (Kitagawa Seiki 6327.T, small-mid cap) — both with replacement difficulty equal-or-higher than VG's and market caps at fractions of a percent of VG's. If you believe the VG thesis, look harder at Fulltech and Kitagawa — higher-leverage plays on the same wave (and ideally Shengyi 600183.SH on premium CCL).

TAM (per GS Jan 2026 revision via STF): AI PCB TAM $26.6B in 2027 (140% CAGR 2025-27, up from $17.4B prior); AI CCL TAM $18.3B in 2027 (178% CAGR, up from $8B). Growth accelerates 2026→2027 (positive second derivative) driven by Vera Rubin midplane/backplane content jumps. VG SAM ~60-75% of AI PCB TAM (~$16-20B by 2027); at 13-15% share, implied 2027 AI-only revenue ~$2.0-3.0B (vs ~$1.5B total revenue FY2024).

Why now: the next 24 months bring two simultaneous step-ups — (1) Vera Rubin 5-7x's VG content per AI GPU at midplane/backplane, and (2) the materials-tier supply squeeze (Nittobo E-glass exit, SABIC PPO outage, premium CCL tightness, lead times 8-14 wks vs 4-6 normal) flows into pricing power for the qualified high-end fabs that can run M9 + Q-Glass. The share-gain story is narrowing to 3-5 fabs that can deliver M9 + Q-Glass + 6+N+6 HDI — and VG is one of them. Risk: Vera Rubin slips, hyperscaler capex normalizes, or VG fails to qualify on next-gen materials — all real, none consensus.

Competitive landscape:

Company	Ticker	Segment(s)	Mkt cap (USD est)	AI mix %	Moat	Pure-play?
Victory Giant	002476.SZ / 2476.HK	AI server PCB pure-play	~$40B	>70%	Yield know-how + share gainer	Yes
Shennan Circuits	002916.SZ	AI server + IC substrate + networking	~$15-18B	40-50%	Diversified Chinese leader	Mostly
Suntak	002815.SZ	AI server + automotive	~$5-7B	~30%	Chinese mid-tier	Yes
Aoshikang	002913.SZ	AI server + automotive	~$4B	~30%	Chinese mid-tier	Yes
Tripod	3044.TW	NVSwitch tray + CSP	~$8B	~30-40%	Taiwan cluster + NVSwitch rumored primary	Mostly
Unimicron	3037.TW	IC substrate + advanced PCB	~$10B	~20-25%	Substrate moat	Diversified
Compeq	2313.TW	HDI + optical	~$3B	~25%	Smaller Taiwan	Yes
Ibiden	4062/4062	4062.T	IC substrate + advanced PCB	~$10B	~20% (PCB)	Substrate dominance
Nan Ya PCB	8046.TW	IC substrate + standard PCB	~$2.5B	~15%	Diversified	Diversified

Moat: (1) yield at high layer count (durable, 12-24mo to replicate); (2) NVIDIA design-in switching cost; (3) geographic decoupling (China+Thailand+Vietnam, only a handful of fabs offer non-China origin at AI quality); (4) capital scale post-HK. Not a moat: patents (PCB tech lightly patented), contractual exclusivity (NVIDIA dual-sources), materials-tier integration (relies on Panasonic/Shengyi for premium CCL). Pricing power moderate — better than commodity fab, weaker than upstream materials; supply tightness largely captured upstream.

Business quality 3-test: (1) 5-year lock-up — yes, at the right price (structurally advantaged business, multi-year secular tailwind; the unhappiness is valuation not quality); (2) unique economic engine — yield at high layer count from 15+ years of process know-how + capital, durable 5-10yrs before photonic/glass-core displacement reshapes the landscape; (3) blank-check disruptor — disruption hard but not impossible (24-36mo to replicate yield know-how, plus a successful NVIDIA qualification cycle). Quality verdict: high-quality, durable, but not unassailable — concentration on NVIDIA + hyperscaler capex cycle means demand-side risk is real. Cycle: mid-2nd inning of an AI capex up-cycle; VG's high AI mix makes it more sensitive to AI-capex moderation than diversified peers — the hedge is hyperscaler custom silicon spreading the base.

Emerging threats: Chinese mid-tier fabs (Suntak, Aoshikang, Wus) climbing the ladder (24-36mo catch-up); photonic interconnect / co-packaged optics (2028+, net PCB-demand effect probably neutral); glass-core packaging substrate (hits Ibiden/Unimicron more than VG).

Management

Governance caveat: SZSE/HKEX disclosure is materially thinner than US SEC — no DEF 14A equivalent with granular related-party / shell-company / executive-comp-by-individual detail. The forensic governance work here is necessarily lighter than the skill template intends; where unverifiable, flagged.

Leadership:

Name	Title	Tenure	Background
Chen Tao (陈涛)	Chairman / Founder	Since founding (2003)	Born 1972 (age ~53); former PLA soldier (left army 1991), then civil servant; sales role at a Taiwanese-owned PCB factory before founding VG. Self-taught entrepreneur, not engineering-trained. Established the HDI division 2019 — the strategic bet that set up the AI-accelerator share gain.
Liu Chunlan (刘春兰)	Spouse / co-controlling shareholder	Since founding	Chen Tao's wife. Per Bloomberg Billionaires Index (Nov 2025), Chen + Liu combined net worth US$9.1B from ~27% combined stake — matches verified Shenghua Xinye 13.92% + HK Victory Giant 13.57% prospectus disclosure.
Chen Bing (陈兵)	CEO / President	Multi-year	Operational lead through the 2015 SZSE IPO and the 2024-26 AI ramp (relationship to Chen Tao not publicly clarified — likely family, unverified)
Various	CFO / VP Finance	n/a	Limited public disclosure; names in HKEX prospectus

(The original 2026-05-03 draft described Chen Tao as "engineering background" — superseded by the verified bio above: ex-PLA + civil servant, self-taught, not engineering-trained.)

Founder-led — generally a positive in the Asian PCB context (long-term capacity decisions, supplier relationships, yield culture). Succession risk read (updated 2026-05-12): Chen Tao is 53, not retirement-age, actively leading through the AI ramp; no public successor signal beyond CEO Chen Bing. No major near-term key-person risk — re-check at 60+. The 2019 HDI decision (six years before AI server demand exploded) is the kind of long-cycle capital allocation that justifies the founder-led premium.

Insider ownership (verified, HK prospectus): Shenghua Xinye (Chen-family) 13.92% + Hong Kong Victory Giant (Chen-family) 13.57% = ~27.5% pre-IPO, controlling in any practical sense given dispersed float. Both subject to standard 6-month post-listing lock-up ending 20 October 2026 — the next material insider-liquidity window.

Cornerstones (37 total, US$997M committed at HK$209.88, 6-mo lock-up to 20 Oct 2026): Yunfeng Capital New Alternative (1,493,700 H shares, 1.56%; Jack Ma-backed), Morgan Stanley, Hillhouse Investment, Mirae Asset, CPE Rosewood (2.92% of H-shares offered), Janchor Partners Pan-Asian (2.92%), Bosera, Sunshine Life, Tianhong, Everbright Wealth, Jump Trading, Luhua Daosheng, + 25 more. Predominantly Chinese institutional + select Western anchors; Yunfeng + Hillhouse + CPE + Janchor are credible long-duration capital. The 37-investor breadth (vs typical 8-15) signals demand strong enough to spread allocations widely. Joint sponsors JPMorgan, China Securities International, GF Securities. Offering: 83.3M H-shares at HK$209.88, ~37% A/H IPO discount (vs typical 5-15%) — the most unusual single number; the +57% debut suggests deliberate pricing for strong post-listing performance over underwriter conservatism.

Capital allocation — grade B+/A- (provisional): aggressive but not reckless; capex tracked the share gain; tariff-hedging logic sound. Revenue grew ~CN¥2B → CN¥10.7B 2015-2024 (5x in 9yrs). APCB (2024) strategic not financial (price undisclosed, ROIC unknowable). Vietnam Bac Ninh ($520M, 2025) greenfield. HK IPO use of proceeds (verified): mainland-China capacity expansion (largest bucket: intelligent manufacturing equipment, automation), new HDI/MLPCB facilities in Thailand + Vietnam, supply-chain resilience. 3-year capacity-add plan: total CN¥30B incremental through 2028, with CN¥10B to offshore (Thailand/Vietnam/Malaysia) — the key capex disclosure for sizing the AI ramp. Reinvestment-heavy; A-share dividend ~0.5-1%, not a yield story. (Earlier draft's "likely deployed to capacity/working capital/R&D" is superseded by the verified breakdown above.)

Compensation & shell-entity scan: individual exec comp not disclosed (China A-share norm); Chen-family ownership is the alignment mechanism. No public allegations of shell-entity asset shuffling, IP licensing through insider entities, or related-party fee extraction; no major short-seller reports as of May 2026 (a meaningful absence — shorts surface fast on suspect Chinese names). Open: full Tianyancha/Qichacha forensic scan needs Chinese-language registry access (blocked — needs paid forensic service or Pink's manual access).

Management DD verdict:

Dimension	Rating	Key Finding
Skin in the game	Green	Founder-led; ~27.5% Chen-family ownership verified
Holdings concentration	Yellow	Net worth dominated by VG; no disclosed supplier/customer cross-holdings
Shell / cross-holdings	Yellow (unverifiable)	No public red flags; limited disclosure depth
Capital allocation	Green	Capex tracked share gain; tariff-hedging strategic; use-of-proceeds verified
Compensation alignment	Yellow	Limited disclosure; founder ownership compensates
Governance quality	Yellow	China A-share + HK dual-listing; not US-grade transparency
Litigation / enforcement	Green (no public flags)	No major regulatory actions
Overall management grade	B+ provisional	Founder-led good signal; disclosure limits forensic depth

Catalysts & risks

Near-term catalysts (0-12 months):

• Q1/Q2/Q3 2026 earnings prints with AI-mix updates. Management FY2026 guide (7 May CEO call): AI revenue >60% of total, high-end AI capacity +50% YoY.
• Vera Rubin volume shipments already started Q2 2026 at 50-55% global supply share — catalyst now shifts from "qualification confirmation" to "ramp slope vs guidance."
• Huizhou Building 10 equipment commissioning Jun-Jul 2026; Building 11 Aug 2026 (combined RMB 40bn annual capacity).
• HK index inclusion (HSCEI, MSCI HK after typical 2-month seasoning).
• Vietnam Bac Ninh first-building commissioning H2 2026.
• A/H spread compression as global investors arbitrage the dual listing.

Medium-term (1-3 years): Vera Rubin VR200 ramp 2H26 (midplane $171-256/GPU); VR300 ramp 2H27 (backplane $781-1,563/GPU, 5x jump); hyperscaler custom-silicon wins (Trainium 3, TPU v7, Maia 300, MTIA); Thailand 2nd facility online (TBC).

Reiterated long-range guidance (7 May 2026 CEO call): 3-year revenue CAGR target ≥50% (unchanged); RMB 100bn output target for 2030 (unchanged — implies ~5x growth off FY2025's CN¥19.3B; aggressive but explicitly reaffirmed).

Secular tailwinds: AI accelerator capex super-cycle (5-10yrs); layer-count escalation; material-grade transition (M-series → M9 → next, each a share-shift event); geographic-decoupling premium; hyperscaler custom-silicon proliferation; glass-cloth/resin tightness flowing to fab pricing power. Leading indicators: NVIDIA capex guidance; hyperscaler quarterly capex; Foxconn/Quanta AI-server shipments; Taiwan/Japan/China advanced-PCB utilization; CCL spot prices.

Risks:

Risk	Likelihood	Mitigants / status
NVIDIA single-customer concentration (>60% FY2025)	Always present	Hyperscaler programs; NVIDIA can't switch easily. Structural — reducible over 24-36mo, never zero.
Hyperscaler capex normalization 2027+	Medium	Dual NVIDIA + hyperscaler exposure; networking diversification. Cycle risk structural in PCB.
Vera Rubin slip / share loss to Tripod or Unimicron	Low (revised down per 7 May call)	VG holds 50-55% global Rubin supply, volume shipping Q2 2026, co-developed from start. Closed on supplier-status leg; residual risk is ramp slope vs guidance.
Material qualification failure (M9 + Q-Glass)	Low-medium	Rubin volume shipments underway implies M9+Q-Glass production-qualified at VG. Largely closed — track yield disclosure Q2-Q3 2026.
Thailand / Vietnam ramp execution	Medium-low (revised down)	Thailand A1 producing qualification boards, yields tracking Huizhou parity in 3-6mo; Vietnam first building live H2 2026 (RMB 11-13bn target).
US 232 / 301 tariff escalation on China-origin	Medium	Thailand + Vietnam capacity already in place (partial hedge).
Chinese mid-tier fab catch-up at layer count	Medium-low (24-36mo)	Yield know-how moat; ongoing R&D. Long-run, no single-action closure.
A/H spread compression cap on HK upside	Medium	A-share has own supply/demand; resolves over 6-12mo as institutional flow normalizes.

Dilution risk: Low. HK IPO already issued ~100M+ new shares at HK$209.88; no convertibles/warrants; 2026 capex ≤RMB 20bn OCF-funded → no near-term equity raise. Key-person risk: Chen Tao (founder/chair) — no near-term concern (age 53, active), re-check at 60+.

Bear case. Vera Rubin slips into 2028+ AND NVIDIA dual-sources Rubin midplane/backplane to Tripod/Unimicron at >40% share AND Chinese mid-tier fabs (Suntak, Aoshikang) qualify 30+L AI boards at >70% yield within 24 months. Downside target ~30% pullback from HK$319 → HK$220-230 (around IPO price). Thesis-invalidation: loss of NVIDIA primary status on the Vera Rubin generation, OR a meaningful Chinese mid-tier fab AI-accelerator qualification win at scale. The concentration risk is the single biggest reason this isn't a Tier-1 hold — but the verified >60% NVIDIA exposure means severe earnings risk if NVIDIA capex moderates 2027+ (high probability VG is much more sensitive than diversified peers; Acid Investments: "if numbers are taken down, the entire space can suddenly be very expensive").

Valuation / DCF

Relative valuation (per Acid Investments / Collyer Bridge):

Peer	FY27 P/E	FY28 P/E
Nanya PCB	44.2x	~20x
Unimicron	31.8x	19x
Ibiden / MEC (Japan)	n/a	~20-mid-20s
Victory Giant (HK class)	lower than Nanya/Unimicron (Acid: "absolutely cheap" on FY28 vs Delton)	likely high-teens / low-20s

Snapshot multiples (HK-class basis): market cap HK$314B (~US$40B); EV HK$283B (~US$36B); P/E (TTM) high (~80-100x on FY2024 NI of CN¥1.15B; ~30-40x on FY2025E NI ~CN¥4.29B); EV/Revenue ~24x FY2024 / ~12-15x FY2025E; FCF yield negative (capex phase); dividend ~0.5-1%. HK-only numbers are partial — total A+H market cap is materially higher; treat as directional, derive precise A+H comps from the prospectus.

Implied expectations: at ~15x EV/Revenue on FY2025E with ~30% YoY growth into FY2026-27, the market implies continued share gain + margin expansion through Vera Rubin. Pricing already includes much of the share-gain story; the Vera Rubin content jump is partially priced. Pricing does not yet fully include (a) Vera Rubin midplane confirmed primary-supplier status, (b) backplane (2H27, 5x content), (c) Vietnam ramp at non-China-origin premium pricing.

Disconnect: fundamentals excellent; valuation rich on absolute terms but cheaper than Taiwanese/Japanese peers on FY27 P/E. Relative valuation is the bull case; absolute valuation is the bear case. No formal DCF — China-listed + broken yfinance SZ data make a clean 3-statement model dependent on a prospectus deep-read; flagged as open homework.

Position sizing (generic guidance, not portfolio-specific): high conviction 2-3% of equity at current levels; aggressive 4-5% on a pullback to HK$240-260; outsized 5-7% only if the bottleneck-tier names (Fulltech, Kitagawa Seiki) are also held — VG should not be the only AI PCB exposure. Scale-in over 3-6 months (HK debut +57% may have absorbed near-term institutional demand; expect 1-2 quarters of consolidation). Triggers — Trim: Vera Rubin design loss / NVIDIA capex guide-down >20% / HK price >HK$420 (~30% above current). Add: Vera Rubin primary-supplier confirmation at midplane / pullback to HK$240-260 / Vietnam plant qualification confirmed. Exit: thesis break (Vera Rubin loss, sustained share decline vs Tripod/Unimicron, Chinese mid-tier scale qualification).

Decision log

2026-05-03 — Initial deep-dive (Register D). Verdict: Tier 2 Tactical, NOT Tier 1 Core. Medium-high conviction on business quality, Medium on entry valuation. Established VG as #1 global AI/HPC PCB (13.8% share 1H25, 10x gain in 12mo), NVIDIA anchor for GB200/GB300, HK secondary listing US$2.6B +57% debut. Cross-checked Collyer Bridge (HK pricing, peer P/E) vs STF Research (GS TAM anchor, Vera Rubin content per GPU, upstream bottleneck names). Key open items flagged for follow-up: customer concentration, Vera Rubin supplier status, Chen Tao succession, quarterly incremental margins. Bear target ~HK$220-230 (~30% pullback). Topic linked: supply-chain-security.

2026-05-12 — Gap-close pass (research-gap list largely resolved):

• HKEX prospectus deep-read — CLOSED. Customer concentration verified at NVIDIA >60% FY2025 / >70% Q1 2025, GB200/GB300 ≥50% — materially higher than the prior 40-55% estimate; raised the concentration-risk framing, downward pressure on conviction. Cornerstones verified (37, $997M, 6-mo lock-up to Oct 20 2026); Chen-family stakes verified (Shenghua Xinye 13.92% + HK Victory Giant 13.57% = ~27.5%); use-of-proceeds + FY2025 revenue (CN¥19.3B) verified.
• NVIDIA Vera Rubin supplier triangulation — CLOSED. Independent confirmation (Oreate AI Blog): Shenghong entered NVIDIA H-series supply 2023, passed "GPU200" certification Q4 2024 → Tier-1 supplier, captured >50% of NVIDIA's market share with Q1 2025 binding rates >70%. Matches the 7 May CEO call within rounding.
• Chen Tao succession/age — CLOSED. Born 1972 (age ~53), ex-PLA + civil servant, founded VG 2003, established HDI 2019; wife Liu Chunlan; combined net worth US$9.1B (Bloomberg Nov 2025) from ~27% stake (matches prospectus). No near-term succession risk; re-check at 60+.
• Quarterly incremental margin — CLOSED. Q1 2026 verified: Rev CN¥5.519B (+28%), NI CN¥1.288B (+40%), net margin 23.3% (vs FY24 10.7% / FY25 ~22% / Q1 25 ~21%), incremental net margin ~30%, OCF +399% (Rubin prepayment working-capital release). Margin expansion accelerating into FY26. Net effect: business-quality leg strengthens; entry-valuation stance unchanged.

2026-05-15 PM — Q1 2026 print re-verified; Chen Tao bio + Rubin Tier-1 triangulation + Yunfeng 1.56% + succession/margin gaps confirmed closed. Status maintained: Tier 2 Tactical, conviction on the business edges higher, entry-valuation unchanged.

Net stance: a high-quality business in the best-positioned tier of the AI server PCB cycle, but a single-customer-program bet (>60% NVIDIA) at a rich post-rerate price. Own it sized below a single-name core, scaled into pullbacks (HK$240-260), paired with upstream bottleneck names (Fulltech, Kitagawa Seiki) — not as the only AI-PCB exposure. Tier 2 Tactical with active management around the catalyst calendar.

Sources

Fragments folded into this canonical page (consolidated 2026-06-02; original archived to _migration-archive/2026-06-02/2476/): 2476-deep-dive.md (the original 2026-05-03 deep-dive — strictly superseded by the updated content already on this page; its sole net-new item, the supply-chain-security topic link, is preserved here).

CEO channel check — sell-side call, 7 May 2026

Sell-side summary of a conversation with VGT's CEO (source archived at KB/raw/transcripts/vgt-ceo-call-2026-05-07.md). Data points are integrated into the sections above; captured here as a discrete primary-source note.

1. Rubin is the lead story — Shenghong holds 50-55% of global Rubin platform supply; co-developed with NVIDIA from the start; volume shipments began Q2 2026 on 44-layer boards; Rubin GM >45% (above blended).
2. Google ramp is the second catalyst — high-end AI PCBs for TPU already in mass production; several billion RMB FY2026 revenue guided; materially softens the >60% NVIDIA framing for FY2026.
3. Capacity — two tracks — Huizhou Bldgs 10 & 11 combined RMB 40bn (Bldg 10 Jun-Jul, Bldg 11 Aug, 2-mo slip); Thailand A1 producing qualification boards (Huizhou parity in 3-6mo); Vietnam first building H2 2026 (RMB 11-13bn); 2026 capex ≤RMB 20bn OCF-funded.
4. Technology lead — 70-layer boards in volume production, 100-layer in reserve; claims only vendor globally at 44-layer scale; HDI 10-stage 30-layer in production, 14-stage 36-layer in pre-development; tracking NVIDIA 2-3 years ahead.
5. Guidance — FY2026 AI revenue >60%, high-end AI capacity +50% YoY; 3-year CAGR ≥50% reiterated; RMB 100bn 2030 output target held.

Channel-check caveat: this is a sell-side summary of management commentary — meaningful but management has incentive to position the story. Treat as one input alongside the HKEX prospectus, Collyer Bridge, STF Research, and SZSE quarterly disclosures. The 50-55% Rubin share is the single highest-impact claim and should keep being independently triangulated (NVIDIA supplier disclosures, Foxconn/Wiwynn build commentary, channel checks).

SEC / filings

VG has no SEC filings (China-listed + HK secondary). Equivalent sources: SZSE annual report (2024) (A-share fundamentals, related-party, board); HKEX prospectus (April 2026) (most comprehensive single document — financials, shareholding, lock-ups, cornerstones, risk factors, related-party, capacity plans); A-share quarterly disclosures (季报).

Research gaps — UPDATED 2026-05-12 follow-up

1. HKEX prospectus deep-read — CLOSED 2026-05-03 (concentration, cornerstones, Chen-family stakes, use-of-proceeds, FY2025 revenue all verified).
2. STF Research "AI Supply Chain: Shortage Intensifying Toward 2027" — Founding Member tier; Pink subscription is Paid only. Pink's decision on tier upgrade.
3. NVIDIA Vera Rubin supplier triangulation — CLOSED 2026-05-12 (Oreate AI Blog independent confirmation, matches CEO call).
4. Chen Tao succession / age — CLOSED 2026-05-12 (born 1972, age ~53, no near-term risk).
5. Quarterly incremental margin analysis — CLOSED 2026-05-12 (Q1 2026 print; ~30% incremental net margin; margin expansion accelerating).
6. Tianyancha / Qichacha shell-entity scan — STILL OPEN. Needs Chinese-language registry login; blocked — needs Pink's manual access or a paid forensic service.
7. Cornerstone individual allocations — PARTIAL CLOSE 2026-05-12 (top 3 verified: CPE Rosewood 2.92%, Janchor 2.92%, Yunfeng 1.56%; remaining 34 not publicly disclosed; not load-bearing).
8. Google TPU revenue triangulation — STILL OPEN. No independent corroboration of VG as primary Google TPU PCB supplier (Broadcom is the TPU silicon designer; GUC and others in the supply chain; VG not named in any public Google TPU supplier disclosure). The "several B RMB FY26 from Google TPU" claim remains CEO-call-only. Cross-check on Q2/Q3 2026 SZSE quarterly customer-concentration disclosure if Google appears as a named top customer.

Still blocked: Item 2 (STF tier — Pink's decision); Item 6 (shell-entity scan — Chinese registry access); Item 8 (Google TPU — primary disclosure not yet public).

Cross-check — Collyer Bridge vs STF Research vs this analysis

Question	Collyer Bridge	STF Research	This analysis
Global PCB leader?	Yes (21 Apr 2026 wrap)	Implied via PCB sector framing	Yes — 13.8% global AI/HPC share 1H25 (SCMP via prospectus)
HK pricing attractive?	Implicit yes ("let quick traders wash out, long-onlys take it higher")	Not addressed	Directionally agree — cheaper than Taiwanese peers on FY27 P/E; absolute valuation rich, concentration real. Tier 2 tactical, not Tier 1 core.
AI PCB TAM trajectory?	Implicit large	GS revision: AI PCB $26.6B / AI CCL $18.3B by 2027 (140%/178% CAGR)	Use STF's GS numbers as anchor — the key quantitative bull-case input
Vera Rubin matters?	Not addressed	Yes — midplane $171-256/GPU, backplane $781-1,563/GPU; Q-Glass+M9 highest bar; demand 2H26/2H27	Use STF data; largest near-term catalyst; makes VG's 2H27 structural not cyclical
Bigger alpha — VG or upstream?	Fulltech / Nittobo bottleneck	Kitagawa Seiki + Taiyo Holdings bottleneck-tier	Strongly agree — named PCB layer priced, bottleneck materials/equipment not; Fulltech, Kitagawa, Taiyo higher-leverage
Concentration risk?	Acid: "if numbers are taken down, the space can suddenly be very expensive"	Not addressed	Verified: NVIDIA >60% FY2025, GB200/GB300 >50%. Higher than prior 40-55% estimate; severe earnings risk if NVIDIA capex moderates 2027+

Divergence from Collyer Bridge: their "consolidate then institutional flow takes it higher" framing understates the concentration risk, which applies more to VG than to diversified peers (Shennan, Unimicron). Divergence from STF: STF hasn't written on VG explicitly in the fetched posts — its bottleneck framing augments rather than displaces the named-fab thesis; the Founding-tier "Shortage Intensifying Toward 2027" post likely covers VG (research gap). Agreement: the two-layers-up bottleneck framing is the high-conviction insight — own VG alongside Fulltech (1815.TW), Kitagawa Seiki (6327.T), and ideally Shengyi (600183.SH). SemiAnalysis cross-check: no SA coverage of VG in the local mirror (~/Dropbox/Wafflebun/KB/wiki/semianalysis/); nearest adjacent is AI-cluster TCO / hyperscaler infra, which doesn't address PCB suppliers individually. Treat as "no SA coverage," not confirmation/disconfirmation.

Related vault pages: ai-server-pcb-primer (companion industry primer) · Union Tool (6278) (PCB drilling consumables) · supply-chain-security

Key external sources: Wikipedia (founding, IPO history, 2024 financials, footprint); SCMP 3350801 (HK debut +57.23%, US$2.6B, 431x retail, 13.8% 1H25 share) + 3349924 (China PCB IPO push); Acid Investments (Quick idea: A/H listing of Victory Giant 2476.HK — peer P/E, bull/bear); Collyer Bridge / @illyquid APAC Wrap 21 Apr 2026 (KB/raw/substack-archive/illyquid/2026-04-26-apac-wrap-21-april-2026.md) + 17 Apr 2026 (...2026-05-02-apac-wrap-17-april-2026-900.md); STF Research — GS AI PCB/CCL TAM (Jan 7 2026, KB/raw/substack-archive/stf-research/2026-01-07-gs-aggressively-revises-up-ai-pcb-ccl-tam.md), Taiyo Holdings solder mask (Jan 31, ...2026-01-31-taiyo-holdings-...md), Kitagawa Seiki (Mar 16, ...2026-03-16-kitagawa-seiki-...md); HKEX prospectus (April 2026); SZSE Q1 2026 filing (29 Apr 2026); IT Business Net (FY2025 NI +273%); Oreate AI Blog (NVIDIA Tier-1 triangulation); Bloomberg Billionaires Index (Nov 2025, Chen+Liu US$9.1B); sell-side CEO channel check 7 May 2026 (KB/raw/transcripts/vgt-ceo-call-2026-05-07.md); Yahoo Finance 2476.HK price only (002476.SZ broken — do not use).

Briefings

• 2026-05-03 · Victory Giant Technology (002476) · vault