最近供应链错配让算力产业链的实际出货节奏总是不如市场预期 —— 比如设备厂说出货量受零部件供应影响，新架构的出货受载板技…
最近供应链错配让算力产业链的实际出货节奏总是不如市场预期 —— 比如设备厂说出货量受零部件供应影响，新架构的出货受载板技术成熟度影响。**怎么预判这类问题要来了？又怎么预判它快结束了？**
Recently, supply chain mismatches have meant that the actual shipment pace of the computing power industry chain consistently falls short of market expectations — for example, equipment makers say shipment volumes are affected by component supply, and shipments of new architectures are affected by the technological maturity of substrates. **How can we anticipate that this kind of problem is about to arrive? And how can we anticipate that it is about to end?**
The desk answers已核实Walnut, from public filings
先把三件事对齐，不然下面全是空谈。
**第一件，Lam 那条说法我在一手文件里没找到。** 泛林 10 月 21 日发的季报新闻稿（截至 2026 年 9 月 27 日的季度），收入 82.2 亿美元、环比增 22%、毛利率 52.3%、每股收益 2.19 美元，CEO Tim Archer 的原话是「对技术拐点和产能的战略投资正在见效，增强了我们跑赢晶圆设备市场的能力」，新闻稿里写的是「创纪录的收入、利润和每股收益」。**整份新闻稿没有一句「零部件供应影响出货」。** 它的 10-Q 里确实有供应风险条款，但那是法律模板，每一季都在，其中还说「稀土这类原材料的获取受限，可能导致延迟、成本上升或无法交付」—— 这句是新的，但你分不清它是新写的事实还是律师补的条款。所以「出货量会受零部件供应影响」这句话，要么出在电话会的问答环节（那个不落到 SEC 文件里，我拿不到逐字稿），要么是媒体转述。**我按一手文件能确认的是：设备这一环现在不缺货，它在满产交付。**
**第二件，Rubin Ultra 和 Vera Rubin 要分开。** 英伟达 8 月 26 日那份 10-Q（截至 2026 年 7 月 26 日）写的是：Vera Rubin 已经在 2027 财年第三季度开始量产出货，往后 Blackwell 和 Rubin 两种系统会同时出货，并且「目前正在经历某些供应约束」。说的是**已量产 + 有约束**，不是「出不来」。Rubin Ultra 是更后面那一代，它的出货计划我没在英伟达自己的文件里找到任何「受载板成熟度影响」的表述。
**第三件，真正被一手文件证实的是载板这一环。** 揖斐电（Ibiden）2026 财年一季度的说明会问答里，几句原话很硬：需求超过行业供给能力的状况「在持续」；在全行业供给短缺的背景下，IC 封装基板本身的均价「涨得比预期多」；所有客户都在询大尺寸高层数基板，他们认为需求会「持续显著超过行业供给能力」。最能说明紧张程度的是一个时点变化 —— **扩产决策的时间从「本财年到下财年之间」提前到了「最迟本财年底」。**
所以这件事的物理形态是：**设备端满产、载板端仍缺、最上游的玻璃布开始松。** 瓶颈不是在消失，是在往产业链的下游移动。
---
这类问题到底是什么
它不是需求问题，是**设计的节奏和供给的物理节奏不同步**。
芯片从设计冻结到量产，要跨七八个环节，每个环节扩产要的时间不一样：
- 玻璃布、铜箔这类最上游材料：十八到二十四个月
- 封装基板：十二到十八个月
- 前道设备：九到十二个月
- 先进封装：六到十二个月
一代产品从图纸到出货大约是两年，而最上游材料扩一次产也要两年。**两个两年不可能刚好对齐。** 所以错配不是偶发事故，是结构性的 —— 每换一代产品就重来一次，只是每次卡在不同的环节。
这解释了为什么它看起来总在「意外」发生：因为大家盯的是需求预测，而卡住的是物理产能。
怎么预判它要来了
按提前量从长到短排，只有前两条是真领先的：
**一、最上游材料的价格和交期。** 玻璃布、铜箔、树脂，这三样是覆铜板和载板的上游。它们涨价、交期拉长，比整机出货出问题早两到四个季度。这条最有用的地方是它骗不了人 —— 材料厂没有理由替下游遮掩。
**二、载板厂「扩产决策的时点」有没有提前。** 这是这次最值钱的信号。揖斐电把决策时点从「本财年到下财年」提前到「最迟本财年底」，等于告诉你：按原来的节奏来不及了。**扩产决策提前，是紧缺加剧的确认，不是缓解的信号** —— 很多人会读反，以为扩产是好事，其实从决策到量产还有一年多，这一年多就是最紧的时候。
**三、设备厂「订单增速」和「收入增速」的缺口。** 订单涨得比收入快，说明接得下造不出。反过来，收入追平订单，说明产能跟上了。泛林这一季收入环比增 22%、毛利率 52.3% 创纪录，说明设备这一环现在不是瓶颈。
**四、客户电话会的措辞变化。** 从「需求强劲」变成「供应受限」「正在分配产能」，这是同步信号，不是领先信号，但它是**确认信号** —— 公司自己承认了，说明前面几条你已经错过了。
顺带说一句：**设计变更公告比财报更早。** 载板层数从 20 层加到 26 层、材料从有机基板换玻璃基板、封装从一种形式换另一种，这些变化一出来，就意味着原来的供应链要重做一遍。
怎么预判它快结束了
这是你真正问的那一半，也是最难的一半。我的答案是**一句话规则**：
**别看「问题解决了」的新闻，看最上游那一段的供给有没有松。**
理由是：整条链是一个串联结构，出货速度等于最慢那一环。所以「快结束了」的唯一可靠标志，是最慢那一环的物理约束在松开 —— 而不是下游宣布了什么。
按这个规则，现在已经有第一个信号了：**揖斐电在同一个问答里说，跟财年初相比，玻璃布的供给预期「似乎有所改善」。** 玻璃布是载板的上游中的上游。它松了，意味着载板产能的天花板开始抬高 —— 但从玻璃布松到载板放量，中间还隔着一年多。
所以要盯的不是一条，是四条依次出现：
1. **最上游材料交期缩短、停止涨价**（玻璃布 —— 已经出现第一个迹象）
2. **载板均价涨势停住**（现在还在「涨得比预期多」，没停）
3. **设备厂毛利率见顶**（泛林 52.3%，已经在历史高位区）
4. **客户的产能承诺停止增长**（英伟达的供给与产能承诺从上季 1190 亿美元涨到 7 月 26 日的 2790 亿美元，其中本财年剩下 920 亿、下财年 870 亿、再下一年 880 亿 —— 这是公司拿真金白银下的判断：紧张要持续到 2028 年。哪天这个数字不再涨了，说明他们认为够了）
四条里出现三条，可以说这一轮错配基本过去了。
但「结束」有两种，必须分清
这是我认为最容易亏钱的地方。
**一种是供给修复型结束**：交期缩短、价格不跌、订单还在。出货放量，业绩兑现，但这时候估值通常已经把这件事反映完了。
**另一种是需求崩塌型结束**：交期也缩短、价格开始跌、订单下修，同时还出现超额库存减值。问题「消失」了，但股价早就跌完了。
区分的判据就一条：**看价格。** 供给修复的时候价格是黏的，需求崩塌的时候价格是第一个松的。
英伟达自己的账上就有两个现成的例子：2026 财年一季度，H20 提了 45 亿美元的超额库存和采购承诺减值；2027 财年上半年，H200 又提了 4 亿美元。**两次都不是「造不出来」，是「备多了」。** 备货过头的代价，和缺货的代价一样真实 —— 这是「错配」的另一面，多数人只盯缺货那一面。
什么情况算我看错
第一，如果玻璃布价格重新开始涨、交期重新拉长，那我说「最上游在松」就是错的，瓶颈还在材料。
第二，如果载板均价在接下来两个季度里停止上涨，而下游订单还在涨，那说明供给追上了，「还要一年多」这个判断偏保守，我低估了扩产速度。
第三，如果我说的「设备端不缺货」错了 —— 也就是泛林下一季收入环比不再增、但订单还在涨 —— 那瓶颈就在设备，不在载板。
第四，也是最重要的：**如果下游出现第二次超额库存减值，那这一轮的性质就从「供给不足」变成「需求下修」，上面整套推断的前提就没了。**
这一半只回答到这里
今天先到这儿。下面这几块我下次补：
- 每个环节具体盯哪个数字、多久看一次（要能变成一张每周填一次的表）
- 载板良率、先进封装良率这类行业内部数据，公开渠道到底能不能反推、怎么反推
- 这套东西落到 A 股是哪几只票、每个环节对应谁（这需要单独做一遍，不能顺口说）
来源：Lam Research《2026 年 9 月季度业绩新闻稿》2026-10-21（8-K 附件 99.1）；Lam Research 2027 财年一季报 10-Q 2026-10-22；NVIDIA 2027 财年二季报 10-Q 2026-08-26（截至 2026-07-26，含供给与产能承诺附表）；揖斐电 FY2026 一季度决算说明会资料与问答 2026 年（en_kessannsetsumei2026Q1 / en_QA_FY2026Q1）；Lam 电话会逐字稿未能取得，相关说法本文未采信。
First, let's align on three things, otherwise everything below is empty talk.
**First, I could not find that Lam statement in the primary documents.** Lam Research's quarterly earnings press release issued on October 21 (for the quarter ended September 27, 2026) reported revenue of 8.22 billion US dollars, up 22% quarter-on-quarter, gross margin of 52.3%, and earnings per share of 2.19 US dollars. CEO Tim Archer's exact words were "strategic investments in technology inflections and capacity are paying off, strengthening our ability to outperform the wafer equipment market," and the press release described "record revenue, profit and earnings per share." **The entire press release does not contain a single sentence saying "component supply affects shipments."** Its 10-Q does contain supply risk clauses, but those are legal boilerplate present every quarter, and they also state that "restricted access to raw materials such as rare earths could result in delays, increased costs or inability to deliver" — this sentence is new, but you cannot tell whether it is a newly written fact or a clause added by lawyers. So the sentence "shipment volumes will be affected by component supply" either came from the Q&A session of the earnings call (which does not go into SEC filings, and I cannot obtain a verbatim transcript), or it was media paraphrase. **What I can confirm from primary documents is: the equipment link is not short of supply right now; it is delivering at full production.**
**Second, Rubin Ultra and Vera Rubin must be separated.** Nvidia's 10-Q dated August 26 (for the period ended July 26, 2026) states: Vera Rubin has already begun volume shipments in the third quarter of fiscal year 2027, and going forward both Blackwell and Rubin systems will ship simultaneously, and it "is currently experiencing certain supply constraints." What this says is **already in volume production + has constraints**, not "cannot come out." Rubin Ultra is a later generation, and regarding its shipment plan I did not find any statement in Nvidia's own documents that it is "affected by substrate maturity."
**Third, what is truly confirmed by primary documents is the substrate link.** In the earnings call Q&A for Ibiden's first quarter of fiscal year 2026, several exact quotes are very firm: the situation where demand exceeds industry supply capacity "is continuing"; against the backdrop of industry-wide supply shortage, the average price of IC packaging substrates themselves "rose more than expected"; all customers are inquiring about large-size, high-layer-count substrates, and they believe demand will "continue to significantly exceed industry supply capacity." What best illustrates the degree of tightness is a change in timing — **the timing of the capacity expansion decision was moved up from "between this fiscal year and next fiscal year" to "by the end of this fiscal year at the latest."**
So the physical form of this matter is: **the equipment end is at full production, the substrate end is still short, and the most upstream glass cloth is starting to loosen.** The bottleneck is not disappearing; it is moving downstream along the supply chain.
---
What exactly is this type of problem
It is not a demand problem; it is **the pace of design and the physical pace of supply being out of sync**.
From design freeze to volume production, a chip crosses seven or eight links, and each link requires a different amount of time to expand capacity:
- Most upstream materials such as glass cloth and copper foil: eighteen to twenty-four months
- Packaging substrates: twelve to eighteen months
- Front-end equipment: nine to twelve months
- Advanced packaging: six to twelve months
It takes about two years for a generation of products to go from drawings to shipment, and expanding capacity for the most upstream materials also takes two years. **Two two-years cannot possibly align exactly.** So the mismatch is not an occasional accident; it is structural — it recurs with every generation change, only each time it gets stuck at a different link.
This explains why it always seems to happen "unexpectedly": because everyone is watching demand forecasts, while what gets stuck is physical capacity.
How to anticipate that it is coming
Ordered by lead time from long to short, only the first two are truly leading:
**One, prices and lead times of the most upstream materials.** Glass cloth, copper foil, resin — these three are upstream of copper-clad laminates and substrates. When they rise in price and lead times lengthen, it is two to four quarters earlier than problems appear in finished-system shipments. The most useful thing about this is that it cannot lie — material suppliers have no reason to cover for downstream.
**Two, whether the "timing of capacity expansion decisions" at substrate makers has moved up.** This is the most valuable signal this time. Ibiden moved the decision timing from "this fiscal year to next fiscal year" up to "by the end of this fiscal year at the latest," which is equivalent to telling you: the original pace is too late. **An earlier capacity expansion decision is confirmation that the shortage is intensifying, not a signal of easing** — many people read it backwards, thinking expansion is good news, but in fact there is still more than a year from decision to volume production, and that year-plus is the tightest period.
**Three, the gap between "order growth" and "revenue growth" at equipment makers.** Orders rising faster than revenue means they can take the business but cannot build it. Conversely, revenue catching up with orders means capacity has caught up. Lam's revenue this quarter rose 22% quarter-on-quarter with a record gross margin of 52.3%, indicating that the equipment link is not the bottleneck right now.
**Four, changes in wording on customer earnings calls.** Shifting from "demand is strong" to "supply is constrained" and "we are allocating capacity" is a coincident signal, not a leading signal, but it is a **confirmation signal** — when the company itself admits it, it means you have already missed the previous items.
Incidentally: **design change announcements come earlier than financial reports.** Substrate layer counts going from 20 layers to 26 layers, materials switching from organic substrates to glass substrates, packaging switching from one form to another — once these changes appear, it means the original supply chain has to be redone.
How to predict that it is about to end
This is the half you actually asked about, and also the hardest half. My answer is a **one-sentence rule**:
**Don't look at news that "the problem is solved"; look at whether supply at the most upstream segment has loosened.**
The reason: the whole chain is a series structure, and shipment speed equals the slowest link. So the only reliable sign that it is "about to end" is that the physical constraint at the slowest link is loosening — not what downstream has announced.
By this rule, the first signal is already here: **Ibiden said in the same Q&A that, compared with the start of the fiscal year, the supply outlook for glass cloth "seems to have improved somewhat."** Glass cloth is the upstream of the upstream of substrates. If it has loosened, it means the ceiling on substrate capacity has started to rise — but from glass cloth loosening to substrates ramping volume, there is still more than a year in between.
So what to watch is not one item but four appearing in sequence:
1. **Lead times for the most upstream materials shorten and price increases stop** (glass cloth — the first sign has already appeared)
2. **The rise in substrate average prices stops** (right now it is still "rising more than expected," not stopped)
3. **Equipment makers' gross margins peak** (Lam Research 52.3%, already in the historical high range)
4. **Customers' capacity commitments stop growing** (NVIDIA's supply and capacity commitments rose from 1190 yi dollars last quarter to 2790 yi dollars as of July 26, of which 920 yi remains for this fiscal year, 870 yi for next fiscal year, and 880 yi for the year after — this is the company putting real money behind its judgment: tightness will last until 2028. The day this number stops rising means they think it is enough)
If three of the four appear, it can be said that this round of mismatch is basically over.
But there are two kinds of "end," and they must be distinguished
This is where I think money is most easily lost.
**One is a supply-repair-type end**: lead times shorten, prices do not fall, orders are still there. Shipments ramp, results are delivered, but by then valuations have usually already reflected this.
**The other is a demand-collapse-type end**: lead times also shorten, prices start to fall, orders are revised down, and excess inventory write-downs appear at the same time. The problem "disappears," but the stock price has long since finished falling.
There is only one criterion for distinguishing them: **look at price.** When supply is repaired, prices are sticky; when demand collapses, price is the first thing to loosen.
NVIDIA's own accounts have two ready examples: in the first quarter of fiscal 2026, H20 took a 45 yi dollar write-down on excess inventory and purchase commitments; in the first half of fiscal 2027, H200 took another 4 yi dollars. **Neither time was it "cannot build it"; both were "stockpiled too much."** The cost of overstocking is as real as the cost of shortage — this is the other side of "mismatch," and most people only watch the shortage side.
What would count as me being wrong
First, if glass cloth prices start rising again and lead times lengthen again, then my saying "the most upstream is loosening" is wrong, and the bottleneck is still materials.
Second, if substrate average prices stop rising in the next two quarters while downstream orders are still rising, that means supply has caught up, the judgment "more than a year to go" is too conservative, and I underestimated the speed of capacity expansion.
Third, if my statement "there is no shortage on the equipment side" is wrong — that is, if Lam Research's next-quarter revenue no longer grows quarter-on-quarter while orders are still rising — then the bottleneck is equipment, not substrates.
Fourth, and most important: **if downstream sees a second excess inventory write-down, then the nature of this round changes from "insufficient supply" to "demand revised down," and the premise of the whole set of inferences above is gone.**
This half is answered only up to here
That is it for today. I will add the following blocks next time:
- Which specific number to watch at each segment, and how often to look (it needs to become a table filled in once a week)
- Whether industry-internal data such as substrate yield and advanced packaging yield can actually be back-derived from public channels, and how to back-derive them
- Which A-share tickers this maps to, and who corresponds to each segment (this needs to be done separately; it cannot be said off the cuff)
Sources: Lam Research "September 2026 Quarter Earnings Press Release" 2026-10-21 (Exhibit 99.1 to Form 8-K); Lam Research fiscal 2027 first-quarter Form 10-Q 2026-10-22; NVIDIA fiscal 2027 second-quarter Form 10-Q 2026-08-26 (as of 2026-07-26, including the supply and capacity commitments schedule); Ibiden FY2026 first-quarter financial results briefing materials and Q&A, 2026 (en_kessannsetsumei2026Q1 / en_QA_FY2026Q1); the Lam earnings call transcript could not be obtained, and related statements were not relied upon in this article.
先把链条摆开,只用单季数,不用累计。最上游是电子布,宏和科技几乎全部收入来自电子级玻璃纤维布:2026 年前三季这块收入 19.54 亿元,平均售价 14.63 元/米,去年同期 4.97 元/米,涨 194.37%;它的原料电子纱进价从 25.80 元/Kg 涨到 72.57 元/Kg,涨 181.28%。售价涨幅略高于原料,差价全落在单季毛利率上:2025 年三季度 34.92%,2026 年二季度 62.03%,三季度 64.48%。往下一层是覆铜板,生益科技二季度单季收入 108.84 亿元、同比增 54.0%,单季毛利率 32.63%,去年同期 28.14%;它三季报要到 10 月 30 日才出,所以这一层只能用二季度加 10 月 17 日的业绩预增公告,不能拿三季度说事。再往下是印制电路板和封装基板:深南电路三季度单季收入 104.78 亿元、同比增 66.29%,单季毛利率 34.85%,去年同期 31.39%;兴森科技的三季报同样没出,10 月 13 日的预告是前三季净利 4.0 到 4.3 亿元、同比增 204% 到 227%,而它对整体收入的描述只有"平稳增长"。
把这几家排在一起对比,规律就出来了。宏和前三季销量 1.3355 亿米;按去年 4.97 元/米的均价反推,去年同期约 1.70 亿米 —— 卖得比去年少了两成,收入却增 131.65%、净利增 435.63%。深南反过来:收入增 66.29%、净利增 94.42%,量在放。兴森最能说明问题:整体收入只是"平稳增长",但里面 CSP 封装基板这一块写的是"量价齐升,订单饱满",光模块 mSAP 基板"在手订单饱满" —— 需求不缺,能交出来的就那么多,利润涨两倍靠的是价和结构,不是量。所以判卡点不看毛利率有多高,看毛利率和收入增速谁在动:价涨而量不动,说明这层产能已经封顶,只能靠提价和改结构过日子;量放而毛利率只小涨,说明这层还有产能,它在吃规模。宏和、兴森是前者,深南是后者。载板良率这种数在财报里看不到,但有个替代读法:良率在爬,第一表现是量先起来、价不一定涨;现在是价先起来、量不动,这个组合更像产能和工艺封着,不像良率在改善。
钱是怎么走的,一句话讲完:电子纱涨价,落到宏和的营业收入(靠单价);宏和提价,落到生益的营业成本(它在预告里写"面对原材料价格急速上涨");生益再往下推,落到深南的存货 —— 深南存货从 51.40 亿元涨到 97.00 亿元、增 88.72%,公司写的原因就是"业务规模增加以及备料影响"。备料增速 88.72% 高过收入增速 53.84%,多出来的三十几个点,是它在替整条链垫库存。同一个动作还落在另外两个科目:应收账款 90.58 亿元、增 72.53%(公司写"业务订单增加影响");经营活动现金流前三季 19.59 亿元、同比降 15.36%,而净利增 77.54%。利润在涨、现金在退,这一轮增长是拿营运资金换的。
所以"预判它要来"我不看交期传闻,看两个直接记录排队的科目。一是合同负债,也就是客户先付的钱:宏和从年初 1,429.97 万元涨到三季度末 6,467.01 万元,四倍多;深南从 1.7290 亿元涨到 3.6295 亿元、增 109.92%,理由写的是"预收货款增加影响"。客户肯把钱压在你这里只有一个原因:怕拿不到货。这个数比涨价公告早一个季度。二是预付设备款和在建工程,它记录的是未来的供给落在哪一层:深南其他非流动资产从 5.02 亿元涨到 24.26 亿元、增 383.67%,写的是"预付工程款及设备款增加";在建工程从 18.51 亿元涨到 46.30 亿元、增 150.17%,是无锡、南通、广州三个新厂。今天付出去的设备款,十二到十八个月后变成产能 —— 哪一层的预付设备款突然暴增,基本就锁定了那一层四到六个季度后不再紧张,而隔壁没扩产的那层,成为下一个卡点。
"快结束"这一半,我把话说窄:不看"问题解决了"的新闻,看卡点层毛利率的环比涨幅 —— 是涨幅,不是毛利率本身。宏和单季毛利率的环比变化:2026 年一季度比 2025 年四季度高 12.18 个百分点,二季度高 6.38 个,三季度高 2.45 个。12.18、6.38、2.45,一条清楚的三阶递减。价格冲击在成熟,但还没有结束。真要拐头,得同时看到三件事:这个环比数变成 0 或负数;宏和的合同负债从 6,467 万元往下掉,说明不用抢了;深南的存货增速掉到收入增速以下,说明不用囤了。三件同时出现,才叫供给修复。现在三件一件都没有。揖斐电在说明会里说玻璃布供给预期"似乎有所改善"——那是预期;宏和三季报的 14.63 元/米、6,467 万元预收款、64.48% 毛利率,是已经发生的事。预期到现在还没在账上出现过,我不拿它当结论。
反方最强的一击在这儿:宏和自己写的收入增长原因是"本期电子布销售单价上涨,特种电子布销量占比增加"。如果销量少两成是因为砍掉低端布、把机台让给特种布,那毛利率从 34.92% 涨到 64.48% 就是产品结构升级,不是供给卡死。真是结构升级,这套卡点判断从根上就错了 —— 价格不会因为扩产而回落,因为它卖的是另一种布。我分辨不了这一层,得说实话:要分开这两种解释,得看同业同期涨没涨,日东纺、台湾德宏这些厂的电子布售价我这一轮没拿到,这个季度的公开数还没出齐。所以我不把它写成结论,只写进下面的证伪条件。能确认的只有一条:宏和售价涨幅 194.37% 略高于原料涨幅 181.28%,价差在扩大。这个事实在两种解释下都成立,所以它不能用来选边。
还有一件容易看漏的事:这一轮受益最集中的那层,恰恰是垫钱最多的那层。深南净利增 77.54%、现金流降 15.36%,存货增 88.72%、应收增 72.53% —— 它的位置是量归它、钱不归它。所以盯出货节奏的时候,中游的现金流会比利润先坏;哪天存货不涨了、应收掉下来,先别高兴,那可能是下游不给账期了。"结束"有两种,旧办法看价格黏不黏,这里补一条更早的:供给修复型结束,先出现的是卡点层也就是宏和的毛利率环比停涨、而中游生益的毛利率环比反而扩大 —— 涨价的钱开始留在中游,不再被上游拿走。需求崩塌型不会这样,它是三层毛利率一起掉、合同负债一起掉。
什么情况下我错。一,宏和四季度单季毛利率环比转负,或跌破 62.48%(64.48% 减两个点),同时售价不再涨 —— "涨幅还在、卡点没松"就错了。二,宏和四季度销量环比回升超过 15%、售价环比持平 —— 那是产能释放而不是结构升级,它比我想的有弹性,上面那个反方解释成立。三,生益 10 月 30 日的三季报,单季毛利率环比涨幅超过宏和 —— 卡点已经从布移到覆铜板,顺序排错了。四,深南四季度存货增速掉到收入增速以下、而应付账款增速仍高于收入 —— 那是下游停止囤货、需求在退,不是供给修复,这一条最会伪装成"错配结束了"。
来源:宏和电子材料科技股份有限公司《2026年第三季度报告》2026-10-23、《2026年第三季度主要经营数据公告》公告编号 2026-087,2026-10-23;广东生益科技股份有限公司《2026年前三季度业绩预增的自愿性披露公告》公告编号 2026—057,2026-10-17;深南电路股份有限公司《2026年三季度报告》2026-10-23;深圳市兴森快捷电路科技股份有限公司《2026年前三季度业绩预告》公告编号 2026-10-052,2026-10-13;揖斐电 FY2026 一季度决算说明会资料与问答。