What this is. Two factories, measured field by field — not a supplier list. Kunshan is the Yangtze-delta counterpart to the Pearl River Delta corridor, and it is where a buyer goes when the requirement is high-layer data-communication board rather than volume FR4. The sample is two because these are the two plants in the city with audited or publicly filed accounts: both are exchange-listed (沪电股份, 002463 since 2010, and 万源通, 920060 since 2024). That fact is the whole method, and it is also the limit: the city hosts many more board plants, and they cannot be measured this precisely.
What the range covers. At the top, the anchor of the cluster: ¥18.945bn revenue, a net margin above 20%, five production bases, 108-layer process capability and 54 layers in volume, a Hong Kong listing application filed in November 2025 — and a Kunshan plant running at 99.7% utilisation. At the bottom, a ¥1.18bn single-site operation that makes single-sided, double-sided and multi-layer board and ships most of it to automotive and appliance customers — 16× smaller, and structurally a different supplier.
What they share is not capability. It is three gaps that survive the size difference entirely: neither publishes a certificate number for any standard it claims; neither publishes the capacity of its Kunshan site (one publishes utilisation percentages, the other publishes capacity for a plant in another province); and neither publishes structured data, so neither can be read by an assistant that is filtering for PCB manufacturing. The smaller one cannot even be reached over a valid HTTPS certificate.
Every row links to a profile built the same way: registry extract, filed accounts, and a direct measurement of the company’s own site on 2026-11-03. Headcount is shown as both figures wherever both exist, because here they never agree.
| Factory | City | Founded | Insured / group | Revenue | What it actually makes |
|---|---|---|---|---|---|
| WUS / Hudian Shares 沪士电子 · 002463.SZ |
Kunshan Yushan |
1992 | 7,052 / 16,512 | ¥18.945bn 2025 |
High-layer data-communication board — backplane to 64 layers, line card 40, server 34, antenna 14; 108-layer process capability, 54 layers in volume; automotive board ¥3.05bn; five bases including Thailand |
| Kunshan Wanyuantong 万源通 · 920060.BJ |
Kunshan Bacheng |
2011 | 234 / 1,409 | ¥1.18bn 2025 |
Single-sided, double-sided and multi-layer board, 12–24 layers named in its own timeline; metal-core, thick copper, ceramic; consumer (game consoles), automotive, industrial control; 150,000 m²/month at Dongtai, not in Kunshan |
Read the revenue column with its year. Both figures are 2025 and both come from filed accounts, so this pair is a cleaner comparison than most in this directory. The insured and group columns are not comparable to each other: one gap is structural (subsidiaries in other provinces and countries), the other is partly disclosed (see below). Neither number is a headcount.
What two listed companies in the same city, 16× apart in size, actually put in public. The pattern is the point: the size difference buys filings and investor records, but it does not buy the two fields a buyer needs most.
| Field | WUS / 沪电股份 | Wanyuantong / 万源通 |
|---|---|---|
| Layer statement | Published per product line: 64 / 40 / 34 / 14 | 12–24 named in a company timeline; no ceiling per line |
| HDI order | 1st and 2nd stage in volume, 3rd sampled, 4th qualifying | 1st and 2nd stage |
| Area capacity | Not published — utilisation only (99.7% Kunshan, 94.4% Huangshi) | 150,000 m²/month — at Dongtai, not Kunshan |
| Certificate numbers | 0 of 8 claimed | 0 of 6 claimed |
| Export share | 81.1% of revenue (2025 H1) | Not stated; customer list only |
| Gross margin | 35.48% | Not stated in the filed fields |
| Insured vs group headcount | 7,052 vs 16,512 — 2.3× | 234 vs 1,409 — 6.0× |
| Structured data (JSON-LD) | 0 blocks | 0 blocks |
| robots.txt / sitemap.xml | 404 / 404 | “File not found” / 302 to a 404 page |
| HTTPS | Valid certificate | Certificate fails validation |
| English site | Full English site at /wus_en/ | None; /en returns 404 |
The revenue gap is not the interesting part; the documentation gap is, and it is much smaller than the revenue gap. For sixteen times the turnover, the larger plant delivers: filed accounts in both Shenzhen and Hong Kong, investor Q&A records that name capacity by site and by month, an industry chapter in a listing document that states its own market share, and a working English site that publishes layer ceilings per product line. That is a real and usable difference, and it is why a high-layer requirement starts at the larger plant.
But the smaller plant delivers something the larger one does not: a plain statement of monthly capacity in square metres — 150,000 m² a month at Dongtai, dated to June 2026. It is the wrong site, but it is a number. The larger plant never gives one for any site; it gives utilisation percentages instead, which tell you the line is full but not how big it is. A buyer sizing a programme needs the area, not the percentage.
The practical reading: money buys audit trails, not disclosure of the fields you need. Both of these companies are among the best-documented board makers in China, and both still leave a buyer asking the same two questions on the first call.
Both profiles show a gap between the insured headcount filed for the Kunshan entity and the group headcount in the consolidated accounts — 7,052 against 16,512 at one plant, 234 against 1,409 at the other. A business database shows only the first number, so a buyer who looks either company up concludes it is smaller than it is. The mechanism behind the gap differs, and the difference matters:
At the smaller plant the gap is structural. The group operates a Dongtai plant and a Thai subsidiary; the insured figure counts the Kunshan legal person. Run the arithmetic and only the group denominator produces a plausible ratio: 1,800,000 m² a year over 1,409 staff is 1,277 m² per person, or 2.2× this directory’s audited benchmark of 572. Over 234 insured it is 7,692 — 13.4×, which no multilayer plant survives.
At the larger plant the gap is partly disclosed. Its Hong Kong listing document states that the company and some subsidiaries did not pay social insurance and housing funds in full for all employees, with possible fines and compulsory payment. So the insured number is not simply a scope artefact there; it is a figure the company itself says is incomplete, attached to a known remediation liability.
The rule for a buyer is the same at both ends: never read the insured figure as a headcount, and ask for the on-register employee count by site — it is the number that predicts whether your programme has people behind it.
This is the finding that makes the cluster page worth writing. The larger plant is the only company in this directory whose English site publishes layer ceilings — a genuinely useful act of disclosure, and exactly what an assistant answering “who in China builds 60-layer backplanes” would want to quote. It cannot be quoted reliably, because there is nothing machine-readable around the numbers: no structured data, no sitemap, no robots file, and a title tag that is the brand name and nothing else.
The smaller plant is at the other extreme: a builder-template page, Chinese only, on a certificate browsers reject, with robots returning “File not found” and a sitemap that redirects to the 404 page. Its capabilities are real — metal-core, thick copper, ceramic, 12–24 layers — and none of them exist in any form a crawler can classify.
Between the two, the cluster covers the whole failure range of Chinese industrial web presence: has the numbers, hides them from machines and does not have the words at all. Neither is visible to the channel its next export customer now uses first.
| Field | Published by | Why it matters to a buyer |
|---|---|---|
| Certificate number, any standard | 0 of 2 | Without a number there is no registry lookup, so “IATF 16949” stays a claim |
| Capacity of the Kunshan site | 0 of 2 | The contracting entity and the producing site cannot be sized |
| MOQ | 0 of 2 | Panel utilisation, and therefore price per piece, stays unknown |
| Lead time on a named stack-up | 0 of 2 | Both plants are capacity-constrained, so this is the field that decides |
| US customs shipment records | 0 of 2 checked | Export claims cannot be corroborated from the buy side |
None of this is an accusation. It is the state of the industry’s public record, and it is why this directory asks rather than rates.
If the requirement is 30 layers and above, high-speed data communication or AI-server class, start at the larger plant — and expect to be told the Kunshan lines are effectively full. If the requirement is single-sided to 12-layer volume in FR4 or metal-core, the smaller plant is the right size of supplier, and a mid-volume order is material to them in a way it is not to the other.
Neither will volunteer it. One has the number internally and publishes utilisation; the other publishes capacity for a different province. This single question separates a real answer from a brochure.
7,052 and 234 are both misleading. Ask for on-register employees by site, and at the larger plant also ask about the social-insurance remediation timetable — it is disclosed, so it can be discussed.
Fourteen standards are claimed across two companies and not one number is published. The certificate check walks the registry lookup; the scope wording decides whether the certificate covers your process at all.
108 layers is a laboratory figure, 54 is the production figure, and the Kunshan line is at 99.7%. Ask at which layer count the quoted yield stops being a sample, and ask what happens to a new programme in a peak quarter.
The larger plant’s English site brands itself “WUS Printed Circuit (Kunshan) Co., Ltd.”, which is not the name of the listed company that signs contracts. Match the USCC on the PO to the plant that ships.
Both profiles were built from the same five inputs: the national business registry (USCC, capital, insured headcount, status), filed accounts and exchange disclosures for revenue and segment data, the company’s own site measured directly on 2026-11-03 (TLS, redirects, robots, sitemap, language paths, document head, structured data), third-party databases for patents and case counts, and — only where a company publishes one — its own English-language capability page.
Where a figure could not be sourced it is left open and named as open, not estimated. The two-plant sample is a limit of method, not a ranking of the city: Kunshan hosts many more board plants, and any of them can be added to this directory on the same fifteen-field standard without changing the standard. No manufacturer paid for inclusion, and none can; corrections with a public source are applied within five working days.