The measurement. Across the sixteen profiles in this directory, compiled 2026-11-01: 5 of 16 publish a layer ceiling as a number — and one of the five publishes it for a factory that has not been built. 10 of 16 publish at least one figure in square metres — seven of them as a capacity claim, the same seven the buyer-question index counts, plus one audited output, one design total and one plant whose two figures disagree; 4 of the 10 publish numbers that contradict each other on their own pages. 6 of 16 publish no area figure at all — and they include three of the four largest plants by revenue. And there is exactly one audited physical output in the whole directory: 2,758,117 m² produced in 2025, from an annual report.
The arithmetic. That audited actual works out to 572 m² per insured employee per year at a plant running 1–108 layers. Set beside it, the loudest stated claim in the directory — 1,000,000 m²/year from a plant with 74 insured employees — implies 24× the audited benchmark. The quietest — 300,000 m²/year from a plant with 293 — implies 1.8×. The spread is the finding.
What this page is. The three kinds of capacity number and how to tell them apart, the full sixteen-row table with sources, the ratio test with its honest limits, and what to ask for instead of a claim.
| Kind | What it describes | How to recognise it |
|---|---|---|
| Audited output | What was actually produced last year, signed by an auditor in an annual report. It includes downtime, yield loss, maintenance, holidays — everything a sales number quietly omits. | Appears in the financial-statements section of a listed company’s annual report, with the production volume next to the revenue. In this directory: one (Champion Asia). |
| Stated capacity | What the equipment could do at some loading the writer found convenient. Not a lie — a different quantity. It answers “what could we sell”, not “what did we make”. | Round numbers on homepages, in catalogues, in municipal press releases. 60,000 m²/month. 300,000 m²/year. Almost always ends in zeros. |
| Designed / target | What a phase under construction is intended to reach. Real money behind it, no boards through it yet. | Watch for the words designed, planned, target, phase 2. Suntak’s headline 2–78 layer ceiling belongs to its unbuilt Thai plant; Zhichao’s 3.5 million m²/year is a design total. |
The three get mixed freely. A supplier quoting “3.5 million square metres” may be reading a design total; a buyer hearing “2–78 layers” may be hearing about a construction site. Neither is dishonest. Both are answers to a different question.
| Plant | Headcount (source) | Layer ceiling | Capacity in m² — kind in brackets | Revenue |
|---|---|---|---|---|
| Champion Asia | 4,824 (2025 year-end, annual report) | 1–108 audited range | 2,758,117 m² produced in 2025 [audited output]; 2,820,933 m² sold; 407,937 m² in stock. The only audited physical output in this directory. | ¥2,535m (net loss) |
| Suntak | 7,174 (registry) | 2–78 Thai plant, unbuilt | Per-factory monthly — the only plant that publishes it: Zhuhai #1 240,000 m²/mo, high-layer plant 120,000, Jiangmen #1 140,000, Jiangmen #2 100,000, Jinzhou 120,000, IC-substrate line 15,000 [stated, per factory]; group total 10.62 million m² of 2025 capacity [stated]. | ¥7,544m |
| Kinwong | 22,278 (registry) | — none published | No square-metre figure anywhere this directory could read [none]. New-capacity story is qualitative: named factories under construction. | ¥15,308m |
| Shengyi Electronics | 6,269 insured (registry: 7,647) | — none published | No area figure on either language version [none]. | ¥9,494m |
| Victory Giant | 17,989 (registry) | — none published | No area figure [none]. Registered as chip manufacturing, not PCB. | ¥10,731m (2024) |
| Yingchuangli / IQE | 1,706 insured | 2–48 stated | No company m² figure; the 15 million m²/year that circulates is the city-wide Suining figure, not the company’s [none]. | “sales above ¥1bn” (own claim) |
| Xunjiexing | 1,515 (registry) | — prototype/volume split instead | 280,000 m²/year design capacity; Zhuhai prototype plant 60,000 m²/mo; Xinfeng #2 “over 1,000,000 m²/year” [stated + designed]. | ¥689m (net loss) |
| Zhichao (Suining) | 1,267 insured | — fine-line HDI, no number | Designed total 3.5 million m²/year after the 2021 phase-2 addition [designed]. | “ceiling above ¥2bn” (stated) |
| Puruisen | 250 insured (claims 500+/700 elsewhere) | 4–8 on product pages; no ceiling | 60,000 m²/month on its own site — and 280,000 m²/year elsewhere on the same site. The two differ by 2.6× [conflict]. | — |
| Jingming | — not published | 2–24 stated | Near 50,000 m²/month as of July 2026 — two months after trial production started; 100,000/mo is a phase-2 target [stated + target]. | — |
| Kaishengshi | 74 insured | up to 10 stated | 1,000,000 m²/year output, own catalogue — from 30,000 m² of plant [stated]. | — |
| Haiying | 293 insured | double-sided through HDI; no ceiling number | 300,000 m²/year (own site); 1.8 million m² produced in 2023 (district government); phase 2 “+60%”. The three cannot all describe the same period [conflict]. | — ¥650M “at full ramp” is a projection |
| Shengda Xinxing | 88 insured | 1–2 stated (single/double-sided aluminium) | ~600,000 m²/year, of which 200,000+ mid-to-high-end aluminium — beside an output value ¥50M (2023) and a revenue ¥43M (2025 database) [three numbers, three measures]. | ¥43m (database) |
| Hong’an Xingsheng | 86 insured | — not published | 1,000,000 m²/year (Baike, post-automation) versus 380,000 m²/year (older recruitment page) [conflict]. | — |
| Glorysky | 844 insured | — not published | No capacity figure in square metres [none]. | ¥1,152m (2022, stale) |
| Wanyexing | 53 insured | — not published | Phase-1 plant is 5,000 m² of floor area — a building, not an output [area ≠ capacity]. | — |
Read the kind column before the number column. One audited figure (green in intent, and it belongs to a plant that lost money making it). A handful of per-factory stated numbers — Suntak’s, which are the best in the directory precisely because they are broken down to a level where a lie would need nine lies. Then conflicts, designs and targets. Six plants publish no area at all, and three of them are the biggest by revenue — which is not an accident and not a scandal, it is an audience: their annual reports answer investors, and buyers are left estimating size from revenue and headcount.
One arithmetic survives every kind of spin, because it uses two numbers a supplier cannot easily inflate in the same direction: annual square metres divided by insured headcount from the latest registry filing. The insured figure is filed to the tax authority; it is the number in this directory that cannot be talked up. The benchmark is the one audited actual:
Benchmark: Champion Asia, 2025, audited. 2,758,117 m² ÷ 4,824 employees = 572 m² per person per year, at a plant running everything from 1 to 108 layers. This is what a real mixed-complexity PCB plant produces per head once yield loss, changeovers, maintenance and holidays are paid for.
| Claim | Implied m² per person / year | × benchmark | What the gap asks |
|---|---|---|---|
| Kaishengshi — 1,000,000 m²/yr, 74 insured | 13,514 | 24× | Either the number describes the group’s ambition, or the m² is mostly single-sided. Both possible — ask which. |
| Shengda Xinxing — 600,000 m²/yr, 88 insured | 6,818 | 12× | Single/double-sided aluminium is the lowest-complexity work there is; a high ratio is expected here. This row is the caveat, in plant form. |
| Haiying — 1.8M m² (gov, 2023), 293 insured | 6,143 | 11× | The government figure and the company’s own 300,000 are 6× apart. Someone is describing a different year or a different site. |
| Zhichao — 3.5M m²/yr designed, 1,267 insured | 2,762 | 4.8× | Designed, not current; fine-line HDI at 5× the mixed benchmark would be exceptional. Ask for current loading. |
| Puruisen — 60,000 m²/mo (720k/yr), 250 insured | 2,880 | 5.0× | Its own other figure (280,000 m²/yr) implies 1,120 — 2.0×, unremarkable. The company publishes both. Ask which plant the first one describes. |
| Haiying — 300,000 m²/yr (own site), 293 insured | 1,024 | 1.8× | Nothing to ask. This is what an honest small-plant claim looks like. |
| Benchmark — Champion Asia audited actual | 572 | 1× | — |
It is complexity-dependent, and the table above shows both directions. Single-sided aluminium legitimately runs many times the mixed benchmark per head — Shengda’s 12× is not a scandal, it is the product mix. High-layer and HDI work runs below it; Suntak, the most transparent plant in the directory, would show a modest ratio on its stated numbers and nobody would call it weak. A second distortion: group figures against single-site headcount, which is probably what Kaishengshi’s 24× and Hong’an’s conflict are describing. And a third: insured headcount excludes dispatch workers, which flatters plants that staff up seasonally.
So the output of the test is never a verdict. It is a ranking of questions: the 24× row gets asked first, the 1.8× row does not need asking. That is what a ratio is for — it tells you where the conversation will earn its keep, before you spend an email on a plant whose numbers already agree.
Kinwong (¥15.3bn, 22,278 people), Shengyi Electronics (¥9.49bn), Victory Giant (¥10.73bn in 2024) and Suntak (¥7.54bn) are, between them, this directory’s entire ¥40bn-plus tier. What they publish in square metres: one set of per-factory figures — Suntak’s. The other three publish layer ceilings? No: they publish no ceiling and no area, on either language version, in the annual-report coverage this directory read. Their sizes are not secret — revenue and headcount are public and audited — but the number a buyer actually needs, the monthly area of the line that would run my order, is not published by any of the four except Suntak, and Suntak publishes it per factory, which is the only useful granularity. Implied revenue per employee puts them at ¥687,000 (Kinwong), ¥1.51 million (Shengyi, on the 6,269 insured floor), ¥597,000 (Victory Giant, 2024) and ¥1.05 million (Suntak) — four numbers that say nothing about capability and everything about where the value sits in each business.
The practical consequence is the reverse of what a buyer expects: the easier a plant is to size up, the smaller it tends to be. Transparency tracks with the need to win individual orders. The plants that no longer need to answer questions have stopped answering them — not deceptively, just completely. Which means the small supplier’s 300,000 m²/year and the giant’s silence are both rational, and the buyer’s only defensible position is to ask every plant the same three questions and compare answers, not claims.
A plant that answers all three in writing is telling you how it plans loading — which is the information the capacity number was supposed to carry. A plant that answers with a bigger round number has also told you something. See which plants let you size your order against them for the index view, and the audit checklist for what to verify when you are standing on the line.
Where this sits: capability by process (the six process pages this measurement undercuts and serves) · high-layer PCB: 5 of 16 state a ceiling (the layer-count half of the claim census) · PCB certificates: claims without numbers (the same method, on the certification layer) · which plants let me size my order against them? · is this plant too big — or too small — for my order? · the Suining cluster, measured (nine of these sixteen in one industrial estate) · what a Chinese PCB quote actually contains · About & methodology.