The market says high-speed copper clad laminate is still raising prices, and board makers cannot get their hands on supply. In Taiwan Union Technology's (2383) Q3 financial report before November 14, which will turn first, gross margin or inventory days, to refute this statement?
The market says high-speed copper clad laminate is still raising prices, and board makers cannot get their hands on supply. In Taiwan Union Technology's (2383) Q3 financial report before November 14, which will turn first, gross margin or inventory days, to refute this statement?
市场说高速覆铜板还在涨价、板厂抢不到货。台光电（2383）11月14日前的三季度财报里，毛利率和存货天数哪个先掉头，就能推翻这句话？
The desk answersverifiedWalnut, from public filings
Both numbers in the third-quarter report are still moving in the tighter direction: quarterly gross margin of 37.70% is a record high, and inventory turnover is 64.3 days, 4 days fewer than the previous quarter. If one number were really to knock down the statement that "high-speed copper clad laminate is still rising in price and board makers cannot get enough supply," the first to move would be inventory days; gross margin would only catch up a quarter later.
No need to wait for November 14
Taiwan Union Technology's consolidated financial report for the third quarter of year 115 was already announced on October 28. Quarterly revenue 608.85 yi, operating gross profit 229.51 yi, after-tax profit 144.99 yi, all quarterly figures, not cumulative.
The company's cumulative gross margin for the first three quarters is 34.47%, 3.2 percentage points lower than the third-quarter standalone 37.70%, entirely because the first quarter was only 29.42%. The quarterly record high is flattened out in the cumulative table; this question can only be answered with quarterly figures.
What is truly not yet out is the downstream board makers. The third-quarter reports of Gold Circuit Electronics and Tripod Technology will not come until the filing deadline of November 14; checking the Market Observation Post System today, both show "no information found." So the statement "board makers cannot get enough supply" can currently only be verified on the CCL side; the board-maker side cannot be verified and requires another ten days.
Which line item and which quarter the price increase lands in
The chain runs like this: boards for AI servers and switches need the lowest-loss grade of material (graded M7, M8, M9 in the industry) → board makers place orders → Taiwan Union Technology, Taiwan Union Technology and ITEQ open capacity → further upstream are glass fiber cloth, copper foil, and resin.
The price increase first lands in the "operating revenue" unit selling price on the income statement, confirmed on shipment, landing in the third quarter of 2026. Raw material price increases land in "operating cost" and only come in after inventory turns over once—Taiwan Union Technology's turnover is 64 days, two months, landing from the fourth quarter to the first quarter of next year.
This is why gross margin is inherently half a beat slow: it records the selling price negotiated two months ago divided by current raw material costs. Inventory days do not have this lag; it records how much goods are in the warehouse on that day and how much was shipped this month.
Inventory days are already trending down
Third-quarter inventory 270.84 yi, only 14.5% more than the previous quarter's 236.64 yi, while revenue grew 28.8%. Goods are selling faster than they are being stocked.
Inventory turnover days 64.3, previous quarter 68.1, same period last year 71.6. Same-quarter accounts receivable 827.04 yi, turnover 122.3 days, no different from the 122 to 130 days level of the past full year.
This 64 days converts back to 2.1 months of usage: third-quarter operating cost 379.34 yi, 270.84 yi in the warehouse, just enough to last more than two months. At the same time, property, plant and equipment grew from 264.56 yi at the end of September last year to 477.48 yi, up 80% year-on-year.
With plant up 80% and depreciation and material stocking both pressing upward, inventory days still being able to fall by 7 days shows that the pull from demand is greater than the push from capacity expansion.
Inventory cannot keep up with shipments, and receivables have not moved—this is the fingerprint of shortage, not the fingerprint of accumulation.
| Quarter | Revenue (yi) | Gross margin | Inventory days | Receivable days |
|---|---|---|---|---|
| 2025Q3 | 251.46 | 30.13 | 71.6 | 129.4 |
| 2025Q4 | 249.34 | 28.58 | 84.7 | 130.2 |
| 2026Q1 | 330.68 | 29.42 | 81.0 | 126.3 |
| 2026Q2 | 472.76 | 33.85 | 68.1 | 124.3 |
| 2026Q3 | 608.85 | 37.70 | 64.3 | 122.3 |
Two peers, one moving in the same direction, one to read carefully
| Company | 2026Q2 gross margin | 2026Q3 gross margin | 2026Q2 inventory days | 2026Q3 inventory days |
|---|---|---|---|---|
| Taiwan Union Technology | 33.85 | 37.70 | 68.1 | 64.3 |
| Taiyao | 29.72 | 36.93 | 90.2 | 86.9 |
| Lianmao | 22.02 | Not reported | 72.7 | Not reported |
Taiwan Union Technology's direction is consistent with Taiwan Union Technology: gross margin 25.15% → 29.72% → 36.93%, inventory days 117.3 → 90.2 → 86.9. Both moving together shows the price increase is an industry-wide CCL matter, not a change in Taiwan Union Technology's product mix alone.
ITEQ must be read carefully. Its third quarter has not yet been filed, so only the second quarter can be seen: gross margin 22.02%, inventory days rising from 66.1 to 72.7. Gross margin is rising and inventory days are also rising—looking only at inventory days would misjudge it as unsellable.
Putting the three together shows a common point: this wave of price increases is industry-wide, and no one has fallen behind. The difference is only in pace; Taiwan Union Technology and Taiwan Union Technology's inventory days are already being compressed, while ITEQ is still in the stage of restocking.
The strongest rebuttal: rising inventory days could also be a good thing
Taiwan Union Technology's inventory days surged to 117.3 in the first quarter this year, and gross margin only jumped from 25.15% to 29.72% the following quarter. Inventory days rising first and gross margin rising later is entirely possible: upstream glass fiber cloth and copper foil are in short supply, CCL makers pull materials early to lock in prices, and goods piling up in the warehouse is a skill, not a burden.
So looking at inventory days alone is not enough to overturn this statement; it must be read together with receivable days. When there is true accumulation, customers also pay slowly, and receivables lengthen in sync; when it is merely material stocking, goods go out and payment is collected as usual, and receivables do not move. Taiwan Union Technology's third-quarter receivables were 122.3 days, unchanged.
Conversely, if one day inventory days return from 64 to 80 while revenue does not fall, the extra is 66 yi of goods—if operating cost is unchanged, inventory would need to increase from 270.84 yi to 337 yi to support that many days; combined with receivable days trending upward, that would mean the demand side has sprung a leak, not material stocking.
Under what circumstances I am wrong
In Q4, gross margin first fell below 33% while inventory days were still within 65 days — that was fiberglass cloth and copper foil costs biting in first — "still raising prices" was false, "can't secure supply" was true, and my ordering of "inventory days turn first" was wrong.
Sources: Taiwan Union Technology (2383) consolidated statements of comprehensive income and consolidated balance sheets for Q3 of ROC year 115, Market Observation Post System, announced 2026-10-28; Taiwan Union Technology (2383) consolidated financial statements for Q2 and Q1 of year 115 and Q4 of year 114, Market Observation Post System; Taiyao (6274) consolidated statements of comprehensive income and consolidated balance sheets for Q3, Q2, and Q1 of ROC year 115, Market Observation Post System, Q3 announced 2026-10-28; Lianmao (6213) consolidated statements of comprehensive income and consolidated balance sheets for Q2 and Q1 of ROC year 115, Market Observation Post System.
三季报里两个数都还在往紧的方向走：单季毛利率 37.70% 是历史新高，存货周转 64.3 天，比上季还少 4 天。真要有一个数先把「高速覆铜板还在涨价、板厂抢不到货」这句话打掉，先动的是存货天数，毛利率只会晚一季跟上来。
不用等十一月十四日
台光电 115 年第 3 季合并财报 10 月 28 日已经公告。单季营收 608.85 亿、营业毛利 229.51 亿、税后 144.99 亿，都是单季数，不是累计。
这家公司前三季累计毛利率 34.47%，比第 3 季单季的 37.70% 低 3.2 个百分点，原因全在第 1 季只有 29.42%。单季新高在累计表里是被摊平的，这题只能用单季数回答。
真正还没出的是下游板厂。金像电、健鼎的三季报要等到 11 月 14 日申报截止前，今天上公开资讯观测站查，两家都是「查无所需资料」。所以「板厂抢不到货」这句，现在只能在 CCL 这一端核；板厂那一端核不了，还要再等十天。
涨价落在哪一行、哪一季
链条是这样走的：AI 服务器和交换机的板子要用最低损耗的那一档材料（业内按 M7、M8、M9 分级）→ 板厂下单 → 台光电、台燿、联茂开产能 → 再往上是玻纤布、铜箔、树脂。
涨价先落在利润表「营业收入」的单位售价上，出货就确认，落在 2026 年第 3 季。原料涨价落在「营业成本」，要等存货转过一圈才进来——台光电一圈 64 天，两个月，落在第 4 季到明年第 1 季。
这就是毛利率天生慢半拍的原因：它记的是两个月前谈的售价，除以现在的原料成本。存货天数没有这个滞后，它记的是当天仓库里有多少货、这个月出了多少货。
存货天数已经在往下走
第 3 季存货 270.84 亿，比上季 236.64 亿只多 14.5%，营收却多了 28.8%。货卖得比进货快。
存货周转天数 64.3 天，上季 68.1 天，去年同期 71.6 天。同季应收帐款 827.04 亿，周转 122.3 天，跟过去一整年 122 到 130 天的水平没差。
这个 64 天换算回来是 2.1 个月的用量：第 3 季营业成本 379.34 亿，仓库里 270.84 亿，刚够撑两个多月。同一时间不动产、厂房及设备从去年 9 月底的 264.56 亿长到 477.48 亿，年增 80%。
在厂房涨八成、折旧和备料都在往上压的节拍下，存货天数还能少 7 天，说明需求这头的拉力比扩产这头的推力大。
存货追不上出货、应收又没动，这是缺货的指纹，不是积压的指纹。
| 季度 | 营收（亿） | 毛利率 | 存货天数 | 应收天数 |
|---|---|---|---|---|
| 2025Q3 | 251.46 | 30.13 | 71.6 | 129.4 |
| 2025Q4 | 249.34 | 28.58 | 84.7 | 130.2 |
| 2026Q1 | 330.68 | 29.42 | 81.0 | 126.3 |
| 2026Q2 | 472.76 | 33.85 | 68.1 | 124.3 |
| 2026Q3 | 608.85 | 37.70 | 64.3 | 122.3 |
同行两家，一家同向，一家要小心读
| 公司 | 2026Q2 毛利率 | 2026Q3 毛利率 | 2026Q2 存货天数 | 2026Q3 存货天数 |
|---|---|---|---|---|
| 台光电 | 33.85 | 37.70 | 68.1 | 64.3 |
| 台燿 | 29.72 | 36.93 | 90.2 | 86.9 |
| 联茂 | 22.02 | 未申报 | 72.7 | 未申报 |
台燿的方向跟台光电一致：毛利率 25.15% → 29.72% → 36.93%，存货天数 117.3 → 90.2 → 86.9。两家一起动，说明涨价是整个 CCL 行业的事，不是台光电一家的产品结构变了。
联茂要小心读。它第 3 季还没申报，只能看第 2 季：毛利率 22.02%，存货天数从 66.1 天升到 72.7 天。毛利率在涨、存货天数也在涨——只盯存货天数，会把它误判成卖不动。
三家放在一起看有个共同点：这一波涨价是全行业的，谁都没掉队。差别只在节奏，台光电和台燿的存货天数已经在压，联茂还在补库存的阶段。
最强的反驳：存货天数往上也可能是好事
台燿今年第 1 季存货天数冲到 117.3 天，毛利率是下一季才从 25.15% 跳到 29.72%。存货天数先升、毛利率后升，完全可能：上游玻纤布、铜箔缺料，CCL 厂提前拉料锁价，货堆在仓库里是本事，不是包袱。
所以单看存货天数不足以推翻这句话，要配应收天数一起看。真积压的时候，客户付钱也慢，应收会同步拉长；只是备料，货出去照样收钱，应收不动。台光电第 3 季应收 122.3 天，没动。
反过来，如果哪天存货天数从 64 天回到 80 天而营收没掉，那多出来的是 66 亿的货——营业成本不变的话，存货要从 270.84 亿加到 337 亿才够这个天数；再配应收天数往上走，就是需求端漏了，不是备料。
什么情况下我错了
第 4 季毛利率先掉到 33% 以下、存货天数还在 65 天以内，那是玻纤布和铜箔的成本先咬进来——「还在涨价」假了、「抢不到货」是真的，我这条「存货天数先掉头」的排序就错了。
- 01台光电（2383）民国 115 年第 3 季合并综合损益表、合并资产负债表，公开资讯观测站，2026-10-28 公告
- 02台光电（2383）115 年第 2 季、第 1 季及 114 年第 4 季合并财务报表，公开资讯观测站
- 03台燿（6274）民国 115 年第 3 季、第 2 季、第 1 季合并综合损益表及合并资产负债表，公开资讯观测站，第 3 季 2026-10-28 公告
- 04联茂（6213）民国 115 年第 2 季、第 1 季合并综合损益表及合并资产负债表，公开资讯观测站
## 一句话里塞了三个命题
售价还在涨、货不够分，是明说的两句；涨价能落到利润里，是没明说的第三句。前两句在需求侧，第三句在成本侧，而存货天数和毛利率恰好一人守一边。
存货天数是时点数：季末仓库里压着多少货，除以这九十天每天出多少，一天算得出来。客户这个月不下单，下个月天数就变。
毛利率是期间平均数：整季确认的收入除以整季确认的成本，而这两笔账的时间点根本不对齐——收入按当下谈的价，成本按一两个月前买的料。所以成本侧的任何变化，都得先在存货里躺一阵，再慢半拍出现在毛利率上。
这也是为什么这道题不能只答一个名字。问「哪个先掉头」，先得问清楚要证伪的是哪一句。
## 这个指标在它自己身上误报过一次
二〇二五年第四季，存货天数从 71.6 天升到 84.7 天，一个季度多压了十三天货。照「天数升就是卖不动」的读法，那一季就该判定需求转弱。结果下一季营收从 249.34 亿（新台币，下同）跳到 330.68 亿，环比涨 32.6%，毛利率也跟着从 28.58% 回到 29.42%。
也就是说，这一轮里存货天数已经向上掉过一次头，掉了十三天，事后证明是备料，不是积压。上游玻纤布和铜箔在缺，覆铜板厂把料提前拉进仓库锁价——在缺料环境里，存货天数上升是本事，不是包袱。
反方最强的一击在这里：一次不算规律。二〇二五年第四季只给了一个样本，赶上了 AI 板急单接住，不代表下一次天数上升还能靠营收救回来。
我认这一半。但样本少只说明这个指标是必要条件、不是充分条件——它单独出现的时候，在台光电自己身上已经误报过一次。所以「哪个先掉头」这个问法的价值已经被打掉一半。
要把它坐实，必须配一个数：营收环比。天数升、营收还在涨，是备料；天数升、营收环比转负，才是需求漏了。两个条件同时成立，才配叫掉头。
```chart
type: line
title: 台光电单季存货周转天数，2025Q4 那次上升是备料
unit: 天
labels: 2025Q3, 2025Q4, 2026Q1, 2026Q2, 2026Q3
values: 71.6, 84.7, 81.0, 68.1, 64.3
```
## 成本那一半，第一个下行窗口就在第四季
季末的存货够撑 64.3 天。按天数摊，第四季卖出去的货，约七成的料是第三季进的。
所以第四季的营业成本，是第一次大面积吃到第三季的料价。上游玻纤布、铜箔如果在第三季涨过而覆铜板售价没有同步再涨，毛利率就会在存货天数完全不动的情况下先掉——这是成本侧的掉头，跟需求一点关系都没有。
量级也差得远。单季毛利率每掉一个百分点，是 6.09 亿毛利（608.85 亿 × 1%）；存货天数每多十天，是 42 亿现金（379.34 亿 ÷ 90 × 10）。毛利率掉两个点等于存货多压三天。分量差这么远，谁也没法替代谁。
判据落在这里：第四季毛利率掉到 33% 以下、存货天数还在 65 天以内，就是成本先咬进来。那时候「还在涨价」是假的，「抢不到货」还是真的。
## 板厂那一端，十天后看两个数
11 月 14 日截止的是板厂，台光电自己的数 10 月 28 日就公告了。现在这个 64 天只能证明覆铜板这一端紧，证明不了板厂在抢。
中间还有个时点差要记住：第三季确认的涨价收入，到板厂那边是变成营业成本的——料进来、做成板、出了货才进损益。板厂毛利率被压，会出现在第四季到明年第一季，比覆铜板厂的毛利率上行晚两个季度。拿板厂毛利率去判断抢不抢货，永远慢半年。
十天后该看板厂的原材料存货天数。真抢不到货，板厂原材料天数会被压到历史低位，同时应付天数被压短——供应商强势，账期是从板厂手里收回去的。反过来，板厂原材料天数在升、应付天数没动，那所谓抢货只是交期长，分配权并不在卖方手里。
而覆铜板这一端的付款条件，一整年没动过：应收天数 122 天，跟过去一年 122 到 130 天的区间完全咬合。这个数可以回算——827.04 亿应收除以单日营收 6.77 亿，得 122.3 天，跟公告口径自洽。
真要缺到抢，账期会先缩短，或者冒出预收款。现在没有。
所以这句话在报表上的位置是：涨价有证据，抢货待验。台光电 37.70%、台燿 36.93%，两家同时往上，售价确实在涨；但付款条件里找不到任何「抢」的痕迹，板厂拿货照样是 122 天的账期。
> 涨价是真的，抢货还没被付款条件证实。
## 缺货是有等级的
第三家同行联茂的读数跟台光电反向：第二季存货天数从 66.1 天升到 72.7 天，毛利率 22.02%，比台光电同季的 33.85% 低 11.8 个百分点。同一季度、同一条链，一家在清库存，一家在补库存。
这一轮的紧不是均匀铺开的。毛利率高的那家天数在降，毛利率低的那家在升——用行业平均去证明「板厂抢不到货」，什么都证明不了，能证明的只有具体料号、具体客户。
## 上行期看不出领先，看出来了也已经晚了
| 指标 | 台光电 Q1→Q3 | 台燿 Q1→Q3 |
|---|---|---|
| 存货天数 | 81.0→64.3（-16.7） | 117.3→86.9（-30.4） |
| 毛利率 | 29.42→37.70（+8.28） | 25.15→36.93（+11.78） |
台燿两个数在同一个窗口一起动，看不出谁领先一季。台光电去年第四季那次也是一起动：天数加十三，毛利率减 1.55 个点，同一季。
所以「存货天数领先毛利率一个季度」这句话，在这两份报表里站不住。唯一确定的滞后在成本那一侧——料进仓库到进损益，隔着 64 天。这个滞后是会计造成的，不是需求造成的，而且它滞后的是成本，不是销量。
含义有点冷：等存货天数确认掉头，毛利率大概率已经掉了一季。想要提前量，只能在月营收和板厂的拉货节奏里找，季报给不了。
## 这轮涨价可能死在自己的扩产上
不动产、厂房及设备从去年九月底的 264.56 亿长到 477.48 亿，一年增八成，砸下去 212.92 亿——相当于单季营收的 35%、单季毛利额的 93%。这条线比前面两个指标都慢，但它决定这轮周期有多长。
它现在的含义只有一个：需求把现有产能吃满了，存货天数才降得下来。等这批在建产能转成可出货产能，缺口自己就闭合了——不需要板厂砍单，不需要需求转弱，存货天数会自己回升，售价会停。
也就是说，最后推翻「还在涨价」的，大概率不是需求，是供给。这条链过去每一轮都是这么结束的。
所以真正该记的日子不是 11 月 14 日，是下一次法说会上的资本开支口径和产能开出节奏。资本开支还在往上加，涨价就还有一两个季度；资本开支开始收，那就是这轮周期的顶部区域了。
## 什么算看错
- 第四季存货天数回到 80 天以上、单季营收环比转负：需求端掉头，整句话作废，涨价和抢货一起结束。
- 第四季毛利率掉到 33% 以下、存货天数还在 65 天以内：只是成本端掉头，作废的是「涨价能变成利润」，抢货那半句还站着。
- 两个方向都没发生，毛利率继续往上、存货天数继续往下：这句话不但没被推翻，还多了一条证据。
来源：台光电（2383）民国 115 年第 3 季合并综合损益表、合并资产负债表，公开资讯观测站，2026-10-28 公告；台光电（2383）115 年第 2 季、第 1 季及 114 年第 4 季、第 3 季合并财务报表，公开资讯观测站；台燿（6274）民国 115 年第 3 季、第 2 季、第 1 季合并综合损益表及合并资产负债表，公开资讯观测站，第 3 季 2026-10-28 公告；联茂（6213）民国 115 年第 2 季合并资产负债表，公开资讯观测站。