What this is. Three board plants within 40 km of each other in Jiangsu — two in Kunshan, one in Suzhou — chosen on a single rule: the accounts are filed or audited, so the numbers can be compared rather than estimated. Between them ¥60.3bn of group revenue, ¥44.9bn of it board. The Pearl River Delta counterpart, on the same method, is eight plants and ¥71.1bn; the city-level detail for two of these plants is in Kunshan, measured.
What the range covers. At one end, a ¥40.1bn group that bought its way into circuits — flexible board, rigid board, and in September 2025 an optical-module maker. At the other, a ¥1.18bn single-site plant 34× smaller that ships single-sided and 12–24-layer board to automotive and appliance customers. In between, the plant that anchors the region’s high-layer work: 108-layer process capability, 54 layers in volume, and a Kunshan line running at 99.7% utilisation.
What they share is not capability — it is the same three blanks. None of the three publishes a certificate number, across eighteen standards claimed between them. None publishes the capacity of the site that would make the board: one gives utilisation percentages, one gives a square-metre figure for a plant in another province, one gives nothing. And none publishes structured data, so no assistant reading the web can classify any of them as a board manufacturer.
Board revenue, not group revenue, is the column that makes them comparable — the largest of the three is only 64% a board company. Every figure below is 2025 and comes from filed accounts; the profile behind each row shows the source line by line.
| Factory | City | Founded | Insured / group | Board revenue | Ownership model |
|---|---|---|---|---|---|
| DSBJ / 东山精密 002384.SZ · 02384.HK |
Suzhou Wuzhong |
1998 | 1,495 / 39,245 | ¥25.620bn 63.85% of group |
Acquired portfolio — Mflex for flexible board, Multek for rigid, optical modules consolidated 2025-09-30; a sheet-metal stamper that bought its way in |
| WUS / 沪电股份 002463.SZ |
Kunshan Yushan |
1992 | 7,052 / 16,512 | ¥18.143bn 95.8% of group |
Organic single-business builder — one product family for 34 years, five bases, Shenzhen-listed 2010, Hong Kong filing lodged 2025-11-28 |
| Kunshan Wanyuantong 万源通 · 920060.BJ |
Kunshan Bacheng |
2011 | 234 / 1,409 | ¥1.18bn 100% of group |
Single-site challenger — one product ladder from single-sided to 24 layers, listed on the Beijing exchange in 2024, expansion capital going to Dongtai and Thailand rather than to Kunshan |
Why board revenue and not revenue. On group revenue the three are ¥40.1bn, ¥18.9bn and ¥1.18bn — a 34× spread that mostly measures how much non-board business each group carries. On board revenue they are ¥25.6bn, ¥18.1bn and ¥1.18bn: a 21× spread, and a much more honest description of who competes with whom for a panel.
The region is usually described as one industrial mass. It is not: the three plants below sit inside the same 40 km and got their capability by three different routes, which decides what they can be asked to do.
Bought. DSBJ started in 1998 as a sheet-metal stamper — its registry record still files it under code 33, metal products, and its former names are 吴县市东山钣金 and 苏州市东山钣金. The flexible-board business came with Mflex, the rigid business with Multek, and the optical-module line with Source Photonics, consolidated on 2025-09-30. The result is a group that can quote a flex circuit, the board it lands on and the module it goes into on one contract — and whose segment table runs on four completely different margin curves, from 5.18% to 36.74%.
Built. WUS has been making one thing since 1992: high-layer data-communication board. Its own English site publishes the ceilings by product line — backplane 64 layers, line card 40, server 34, antenna 14 — against a process capability of 108 layers and volume production at 54 and above. That is the most usable capability disclosure of any plant in this directory, and it belongs to a company that also reports 99.7% utilisation on the Kunshan lines, which is the answer a buyer gets when they ask for a new programme slot.
Scaled. Wanyuantong is the one that did not buy and did not specialise: single-sided, double-sided and 12–24-layer FR4, plus metal-core, ceramic and thick copper, sold into automotive, appliance and game-console work. It is the smallest of the three by a factor of 21 and the only one that publishes monthly capacity in square metres — 150,000 m² at Dongtai, of which 10,000 is HDI — for a plant that is not in Kunshan.
The practical consequence is a shortlist rule rather than a ranking: if the requirement is layer count, only one of the three is in the conversation; if it is volume FR4 or metal-core at mid-size, only one is the right size of supplier; if it is flex plus rigid plus assembly, only one can contract it as one order.
Two of the three disclose a board-line gross margin, and they are 2.1× apart. Both make multilayer board in the same province from the same laminate supply chain, so the gap is not process — it is what the board is for.
| Plant | Board line | 2025 gross margin | Group margin, same year | Latest reading |
|---|---|---|---|---|
| DSBJ | Electronic circuits (PCB, FPC, rigid-flex) | 17.59% | 14.09% | Group 21.28% for 9M 2026, after optical modules reached a fifth of revenue |
| WUS | PCB segment | 36.91% | 35.48% | 2025 full year; Kunshan utilisation 99.7% |
| Wanyuantong | All board | — | — | Not stated in the filed fields; 2026 H1 revenue ¥594m (+9.88%) |
The 19-point gap between 17.59% and 36.91% is the price of the end market, not of the factory. WUS sells 81.1% of its revenue abroad into switches, routers and AI servers — products where a 30-plus-layer backplane is a small fraction of the bill of materials and a qualification cycle is measured in quarters. DSBJ’s board business is consumer and display driven, where the board is a visible cost line and the customer reprices every season.
DSBJ is now closing the gap from the other direction: its group margin reached 21.28% for the first nine months of 2026, four points above its own board line in 2025, because the highest-margin business in the group is no longer board at all. That is the trap in reading a group margin as a board price — at both companies the group number and the line number differ, in opposite directions.
The one ratio that can be computed across all three, since all three publish both figures. It is a rough instrument — group headcount against board revenue, so a group with large non-board businesses is penalised — but it kills the assumption that scale means efficiency.
| Plant | Board revenue | Group employees | Board revenue per employee |
|---|---|---|---|
| DSBJ | ¥25.620bn | 39,245 | ¥653,000 |
| WUS | ¥18.143bn | 16,512 | ¥1,098,800 |
| Wanyuantong | ¥1.18bn | 1,409 | ¥837,500 |
The smallest plant is not the least productive per head; the largest group is. The 1.7× gap between WUS and DSBJ is layer count and end market again — a 54-layer backplane carries more revenue per square metre and per operator than a display board — and the middle position of the ¥1.18bn plant is a reminder that this ratio measures product mix far more than it measures management.
What it cannot tell you is capacity, because none of the three publishes area for the site that would run your programme. That is the next table.
Same method as every cluster page in this directory: what each company actually puts in public, measured on 2026-11-03. The size difference buys filings, not the fields a buyer needs.
| Field | DSBJ | WUS / 沪电 | Wanyuantong |
|---|---|---|---|
| Layer statement | Not on the site; 78+ layers appears only in broker coverage of the 2025 results | Per product line: 64 / 40 / 34 / 14; 108 capability, 54 in volume | 12–24 named in a company timeline; no ceiling per line |
| Area capacity, own site | None, in either language | None — utilisation only (99.7% Kunshan, 94.4% Huangshi) | 150,000 m²/month — at Dongtai, not Kunshan |
| Certificate numbers | 0 of 4 claimed | 0 of 8 claimed | 0 of 6 claimed |
| Export share | 83.20% of revenue (2026 H1) | 81.1% of revenue (2025 H1) | Not stated; customer list only |
| Insured vs group headcount | 1,495 vs 39,245 — 26× | 7,052 vs 16,512 — 2.3× | 234 vs 1,409 — 6.0× |
| Industry code on the registry | 33 metal products on one record, C3982 on another | C3982, matching the business | Matching |
| Structured data (JSON-LD) | 0 blocks | 0 blocks | 0 blocks |
| robots.txt | HTTP 200, 0 bytes | 404 | “File not found” |
| sitemap.xml | 404 | 404 | 302 to a 404 page |
| HTTPS | Valid — certificate is *.mflex.com, the acquired subsidiary’s wildcard | Valid | Certificate fails validation |
| English site | /en serves the Chinese page, declaring lang="zh-CN" | Full English site at /wus_en/ | None; /en 404 |
| hreflang | None | None | None |
The region’s fourth large board group is in Wuxi, 60 km north-west, and it cannot be measured the way the other three were because its domain does not resolve from mainland China. Measured on 2026-11-03: www.tripod-tech.com returns SERVFAIL from the local resolver, HTTPS times out at 9.9 seconds with DNS resolution failing and no TCP connection, and DNS-over-HTTPS to a public resolver is also blocked from this network. No page, no robots.txt, no certificate, no language check: every digital field in this directory’s fifteen is open for that plant.
Tripod (健鼎, 3044.TW) is a Taiwan-listed group with bases in Taoyuan Pingzhen, Wuxi, Hubei Xiantao and Vietnam, and it is not a marginal company: 2025 consolidated revenue NT$73.399bn (+11.54%), gross profit NT$18.992bn (+24.24%), net profit NT$10.225bn (+38.29%), gross margin 25.88%, EPS NT$19.45 — seven consecutive years above one share of capital in earnings. Its Vietnam Chuc Duc plant is scheduled for trial production in the first quarter of 2026 with roughly 300,000 sq ft a month of new capacity, aimed at notebook and server board.
Two things to take from this, and one caution. First, the figures above are in New Taiwan dollars from Taiwanese filings and news coverage, not RMB from a mainland filing — do not stack them against the ¥ column above without converting. Second, a plant that a buyer in Shanghai cannot resolve is a plant that a buyer’s assistant cannot cite; for a group of this size the answer is not a marketing problem but an infrastructure one. The caution: this is a reachability measurement from one network on one day, not a statement about the company, and no profile will be published here until the site can be fetched and the fifteen fields measured directly.
The three measured plants are not uniformly bad; they fail in three distinct ways, and the distinction matters if you are the one deciding what to fix.
A site that hides what it has. WUS publishes the layer ceilings a buyer wants, on an English site that works, and then surrounds them with nothing a machine can use: no sitemap, no robots file, no structured data, no hreflang, and a title tag that is the brand and nothing else. The capability is real and the disclosure is real; it is simply invisible to the channel that would quote it.
A site that answers the wrong question. DSBJ’s obvious English path, /en, returns the Chinese homepage byte for byte, with lang="zh-CN". A real English section exists — three directories away, under a capital-E /En/Index/listView/catid/… path — and every page in it also declares itself Chinese. Its certificate is the acquired subsidiary’s wildcard. Six days before the measurement it listed in Hong Kong, raising capital from an audience that would read the English site first.
A site that does not load. Wanyuantong sits on a builder template, Chinese only, behind a certificate browsers reject, with robots returning “File not found” and a sitemap that redirects to the 404 page. Its capability claims — metal-core, ceramic, thick copper, 12–24 layers — are exactly the kind of long-tail work a buyer searches for, and none of them exist in a form anything can index.
Across the region: 0 of 3 publish JSON-LD, 0 of 3 publish hreflang, 0 of 3 have a working sitemap, and 1 of 3 has a usable English site. That English site belongs to the plant whose product needs it least in principle — high-layer data-communication board is sold by qualification, not by search — and it is the only one of the three that could be quoted by an assistant asked who in China builds 60-layer backplanes.
| Field | Published by | Why it decides the order |
|---|---|---|
| Certificate number, any standard | 0 of 3 | Without a number there is no registry lookup, so “IATF 16949” stays a claim — at the plant that needs it most, the automotive one |
| Capacity of the site that ships | 0 of 3 | The contracting entity and the producing site cannot be sized, so a lead time cannot be tested against a capacity |
| MOQ | 0 of 3 | Panel utilisation, and therefore the price per piece, stays unknown |
| Lead time on a named stack-up | 0 of 3 | All three are capacity-constrained in the lines a new programme would use |
| US customs shipment records | 0 of 3 checked | Two plants book four-fifths of revenue abroad and neither can be corroborated from the buy side |
None of this is an accusation. It is the state of the industry’s public record at the top of the market, and it is why this directory asks rather than rates.
The three plants are 40 km apart and not interchangeable. Bought capability means one contract across flex, rigid and module — and four margin curves to price against. Built capability means a published layer ceiling and a line that is 99.7% full. Scaled capability means a mid-size supplier for whom your order is material. Start from the requirement, not the map.
At DSBJ the board line earned 17.59% while the group earned 14.09%; at WUS the PCB segment earned 36.91% against a 35.48% group. Whichever direction it runs, the group number is not the price of your panel. Ask which line the quote is costed on, and what margin band it assumes.
¥653,000 of board revenue per employee at the ¥40bn group, ¥1.10m at the ¥18.9bn plant, ¥838,000 at the ¥1.18bn one. The ratio follows product mix, so use it to understand what a plant sells, not how well it is run.
The insured-to-group gaps here run 26×, 2.3× and 6.0×. In every case the site that makes the board is a different legal person from the one a database shows. Put the USCC on the PO and get the shipping address in writing.
Eighteen standards are claimed across three plants and zero numbers are published. The certificate check walks the registry lookup; the scope wording decides whether the certificate covers your process at all — “production of printed circuit boards” and “sales of electronic products” are different certificates.
Two of the three cannot be classified by a machine at all, and the Wuxi group in the section above cannot be resolved from inside China. The plants that win the AI-sourced enquiry will not be the largest ones — they will be the ones whose layer ceilings, capacity and certificate numbers exist in a form a crawler can read.
Three plants selected on one criterion: filed or audited accounts. Registry extracts, 2025 annual reports and the 2026 interim and third-quarter filings for DSBJ; 2025 annual report and Hong Kong listing document for WUS; Beijing Stock Exchange filings for Wanyuantong. Site measurements made directly on 2026-11-03 — DNS, TLS chain, robots.txt, sitemap.xml, the English and Chinese paths, the lang attribute on each, document head, heading count, structured-data blocks. The Wuxi section is built from Taiwanese filings and news coverage only, and says so; no profile exists for that plant and none will be published until its site can be fetched and measured like the others.
Board revenue is taken from the segment table where one is published — DSBJ electronic circuits, WUS PCB segment — and from total revenue where the company is a single business. Per-employee ratios divide board revenue by group headcount and are labelled as such, because no plant here publishes a site headcount. Nothing on this page is estimated, projected or rounded up; where a field is absent it is shown as absent. Manufacturers cannot pay for inclusion in this directory and none has. Corrections with a public source are applied within five working days.
Read next. DSBJ, field by field — three headcounts 26× apart and an English site that declares itself Chinese; WUS / 沪电股份 — the only plant in this directory whose English site publishes layer ceilings; certificates without numbers; stated capacity versus audited output.