In early November, Customs published the October import value of semiconductor equipment, and NAURA's Q3 report has already stated its contract liabilities. Imports are going down while contract liabilities are going up — does this count as domestic equipment picking up the share ceded by overseas players?
In early November, Customs published the October import value of semiconductor equipment, and NAURA's Q3 report has already stated its contract liabilities. Imports are going down while contract liabilities are going up — does this count as domestic equipment picking up the share ceded by overseas players?
海关11月上旬公布10月半导体设备进口额，北方华创三季报的合同负债也已经写明。进口往下、合同负债往上，这算国产设备接住了海外让出的份额吗？
The desk answersverifiedWalnut, from public filings
Half of the premise behind this does not hold: October customs data still has not been released. What the General Administration of Customs discussed at its October 14 press conference was still the first three quarters; the single month of October, by convention, has to wait until mid-to-late November. So "imports heading down" is for now an expectation, not a number that can be checked. What can be checked is the third-quarter report, and the third-quarter report supports only a weaker conclusion: volume was grabbed, but price was not held.
Contract liabilities rose 12.7%, but the work taken in did not increase
On the balance sheet in the third-quarter report, contract liabilities were 53.01 yi yuan, 12.7% more than 47.04 yi yuan in the same period last year and 23.5% more than 42.91 yi yuan at the start of the year. Looking at this number alone, "up" holds.
The problem is that it rose more slowly than revenue. First-three-quarter operating revenue was 314.67 yi yuan, up 15.26% year-on-year. Dividing contract liabilities by trailing four-quarter revenue gives 12.2% now, 12.8% a year ago, and 28.0% two years ago (contract liabilities of 77.83 yi yuan at the end of the third quarter of 2024). This "reservoir" has been shrinking relative to overall size; the 12.7% year-on-year growth is just a bounce from a very low base.
On the same statement, inventory was 339.0 yi yuan, 18.4% more than at the start of the year and 12.2% more than in the same period last year. Converted using single-quarter operating cost of 69.72 yi yuan, inventory turnover is 437 days, versus 408 days in the same period last year. Goods are piling up in the warehouse, and shipments have not kept up.
This mismatch had already weighed on the single-quarter figures by the third quarter: third-quarter single-quarter revenue was 113.06 yi yuan, up only 1.31% year-on-year, while net profit attributable to the parent was 20.22 yi yuan, up 5.18%. Profit did not come from selling more.
The company's own two sentences are clearer than any paraphrase
Affected by tight material supply for individual categories, the company's delivery of some orders faces certain pressure in the short term.
The company provided certain commercial concessions to some large-volume procurement customers.
The first sentence corresponds to the revenue side: orders exist, materials are stuck, acceptance is slow, so contract liabilities remain on the books. The second sentence corresponds to the gross margin side: single-quarter gross margin was 38.33%, 1.98 percentage points lower than 40.31% in the third quarter of last year and 1.0 percentage point lower than 39.33% in the second quarter of this year (single-quarter figures are derived by subtracting cumulative figures; cumulative gross margin fell from 41.41% to 39.44%, as the company itself wrote in the third-quarter report).
Share comes at a price: price concessions to large customers and amortization of validation costs for new products. This is not called catching it; this is called buying it in. What actually catching it looks like is volume up and price stable, both happening at the same time.
Put six companies side by side, and contract liabilities are not an industry indicator at all
| Company | Contract liabilities (yi) | vs. beginning of year | Year-on-year | Q3 single-quarter revenue year-on-year |
|---|---|---|---|---|
| NAURA | 53.01 | +23.5% | +12.7% | +1.3% |
| AMEC | 22.64 | -25.6% | -48.4% | +17.2% |
| Piotech | 45.18 | -6.9% | -7.7% | +1.8% |
| ACM Shanghai | 8.49 | +15.1% | +23.2% | +7.6% |
| Hwatsing | 23.99 | +24.2% | +58.7% | +23.7% |
| Kingsemi | 9.18 | +33.5% | +14.4% | +55.6% |
Year-on-year, four rose and two fell, but looking at quarter-on-quarter, the three largest by revenue scale—Naura, AMEC, and Piotech—all had contract liabilities moving down (-0.3%, -18.3%, and -11.9%, respectively).
AMEC's contract liabilities nearly halved year-on-year, while single-quarter revenue still rose 17.2%, indicating that it delivered faster than it collected; Naura was the reverse, with revenue up only 1.3% and inventory piling up, meaning delivery was slower than collection.
The same line item reads as opposite operating states, because contract liabilities are a net amount: how much money was received minus how much revenue was recognized. Once the customer prepayment ratio changes, it decouples from newly signed orders. Using it to prove "share transfer" is equivalent to using a number rewritten by settlement terms as evidence.
This half-sentence on imports currently has only annual figures
In the United Nations Comtrade database (reported by China Customs), HS848620 (machines for the manufacture of semiconductor devices or integrated circuits) imports in 2023 were 27,408 million US dollars and 11,878 units; imports in 2025 were 34,666 million US dollars and 10,806 units; the 2024 column has no data in the database.
Units fell 9.0%, value rose 26.5%, and unit price rose from 2.31 million US dollars to 3.21 million US dollars, up 39%. This set of numbers does not say "imports are heading down"; it says "imports are concentrating toward high-end": the low-to-mid end has been replaced, and the categories that are bottlenecks still have to be bought, and at higher prices.
So when October data actually comes out, three things must first be separated: whether what fell was units or value; how much fell from each of the three sources—the Netherlands, Japan, and the United States; and whether what fell was the categories domestic suppliers can already make.
There are three causes of a decline in import value, and their implications for the financial statements are completely different. If wafer fab capital expenditure contracts, domestic orders will fall along with it; if controls make purchases impossible, domestic suppliers have orders but delivery is stuck; only genuine substitution will make domestic volume and price rise together.
Which line items this chain lands on
Import value (US dollars, published with a one-month lag) → wafer fab equipment procurement budget → equipment makers' newly signed orders (reflected in contract liabilities and inventory, generally leading revenue by two to four quarters) → customer acceptance and revenue recognition → gross margin.
Customs data sits at the very front of the chain; the third-quarter report sits at the third and fourth links. Matching the two ends directly skips the two intermediate variables of capital expenditure and delivery, so the conclusion is bound to be fragile. To see whether it has been "caught," what should be compared is domestic revenue growth against the decline in imports, the direction of domestic gross margin, and structural changes in import unit prices.
Under what circumstances would I be wrong: if the year-on-year decline in October import value exceeds 20%, and at the same time NAURA's fourth-quarter single-quarter revenue growth returns to above 15% and single-quarter gross margin returns to above 40%, then it has truly caught on, and caught on with pricing power. Conversely, if in the fourth quarter contract liabilities are still hovering around 5.3 billion yuan and single-quarter revenue is still in the single digits, then "contract liabilities going up" is merely a mirror image of slow delivery.
Source: NAURA Technology Group Co., Ltd. "2026 Third Quarter Report" 2026-10-30 Announcement No. 2026-060; Advanced Micro-Fabrication Equipment Inc. (Shanghai) "2026 Third Quarter Report" 2026-10-31 CNINFO ID 1225646685; Piotech Inc. "2026 Third Quarter Report" 2026-10-29 CNINFO ID 1225636861; ACM Research (Shanghai), Inc. "2026 Third Quarter Report" 2026-10-29 CNINFO ID 1225635956; Hwatsing Technology Co., Ltd. "2026 Third Quarter Report" 2026-10-31 CNINFO ID 1225648256; Kingsemi Co., Ltd. "2026 Third Quarter Report" 2026-10-30 CNINFO ID 1225642763; United Nations Commodity Trade Statistics Database UN Comtrade, China Customs-reported HS848620 import value and quantity (2023, 2025 annual).
这条的前提有一半不成立：10月的海关数到现在还没发。海关总署10月14日那场发布会讲的还是前三季度，10月单月按惯例要等到11月上中旬。所以"进口往下"眼下是一个预期，不是一个能核对的数。能核的是三季报，而三季报只支持一个更弱的结论：量是抢到了，价没守住。
合同负债涨了12.7%，收进来的活儿却没变多
三季报资产负债表上，合同负债53.01亿元，比去年同期的47.04亿元多12.7%，比年初的42.91亿元多23.5%。单看这一个数，"往上"成立。
问题是它涨得比收入慢。前三季营业收入314.67亿元，同比增15.26%。拿合同负债除以滚动四个季度的收入，现在是12.2%，一年前是12.8%，两年前（2024年三季度末合同负债77.83亿元）是28.0%。这个"蓄水池"相对体量一直在缩；12.7%的同比增长，只是从一个很低的基数上弹了一下。
同一张表里，存货339.0亿元，比年初多18.4%、比去年同期多12.2%。按单季营业成本69.72亿元折算，存货周转437天，去年同期408天。货在往库里堆，出货没跟上。
这个错位到三季度已经压到单季数上：三季度单季收入113.06亿元，同比只增1.31%，而归母净利润20.22亿元增5.18%。利润不是靠多卖出来的。
公司自己那两句话，比任何转述都清楚
受个别品类物料供应紧张的影响，短期内公司部分订单交付存在一定压力。
公司为部分大批量采购客户提供了一定的商务优惠。
第一句对应收入端：订单有，物料卡住，验收慢，所以合同负债留在账上。第二句对应毛利端：单季毛利率38.33%，比去年三季度的40.31%低1.98个百分点，比今年二季度的39.33%低1.0个百分点（单季数由累计数相减得出，累计毛利率从41.41%降到39.44%，公司自己写在三季报里）。
份额是有对价的：给大客户让价、给新产品摊销验证成本。这不叫接住，这叫买进来。真接住的样子是量往上、价稳住，两样同时发生。
六家摆一起，合同负债根本不是行业指标
| 公司 | 合同负债(亿) | 较年初 | 同比 | Q3单季收入同比 |
|---|---|---|---|---|
| 北方华创 | 53.01 | +23.5% | +12.7% | +1.3% |
| 中微公司 | 22.64 | -25.6% | -48.4% | +17.2% |
| 拓荆科技 | 45.18 | -6.9% | -7.7% | +1.8% |
| 盛美上海 | 8.49 | +15.1% | +23.2% | +7.6% |
| 华海清科 | 23.99 | +24.2% | +58.7% | +23.7% |
| 芯源微 | 9.18 | +33.5% | +14.4% | +55.6% |
同比是四家涨、两家跌，可一看环比，收入规模最大的三家——北方华创、中微、拓荆——合同负债全在往下（分别-0.3%、-18.3%、-11.9%）。
中微合同负债同比几乎腰斩，单季收入还增17.2%，说明它交付快于收款；北方华创反过来，收入只增1.3%、存货堆起来，是交付慢于收款。
同一个科目读出相反的运营状态，因为合同负债是净额：收了多少钱减掉确认了多少收入。客户预付款比例一变，它就跟新签订单脱钩。拿它证明"份额转移"，等于拿一个被结算条款改写的数当证据。
进口这半句，现在只有年度数
联合国商品贸易统计库（中国海关报送）里，HS848620（半导体器件或集成电路制造设备）2023年进口274.08亿美元、11878台；2025年进口346.66亿美元、10806台，2024年这一栏在库里没有数据。
台数少了9.0%，金额多了26.5%，单价从231万美元涨到321万美元，涨39%。这组数说的不是"进口在往下"，是"进口在往高端集中"：中低端被替掉了，卡脖子的那几类还得买，而且更贵。
所以10月数据真出来，也得先拆三件事：掉的是台数还是金额；从荷兰、日本、美国三个来源地各掉多少；掉的是不是国产已经能做的那几类。
进口额跌有三种成因，对报表的含义完全不同。晶圆厂资本开支收缩，国产会跟着掉单；管制买不到，国产有订单但交付卡住；只有真替代，才会让国产的量价一起往上。
这条链落在哪几个科目上
进口额（美元、滞后一个月公布）→ 晶圆厂设备采购预算 → 设备厂新签订单（体现为合同负债和存货，一般领先收入两到四个季度）→ 客户验收确认收入 → 毛利率。
海关数站在链条最前面，三季报站在第三、第四环。把两端直接对上，中间跳过了资本开支和交付两个变量，结论一定是脆的。要看"接没接住"，该比的是国产收入增速对进口降幅、国产毛利率的方向、以及进口单价的结构变化。
什么情况下我错了：如果10月进口额同比降幅超过两成，同时北方华创四季度单季收入增速回到15%以上、单季毛利率回到40%以上，那就是真接住了，而且是带着定价权接的。反过来，如果四季度合同负债还在53亿元附近打转、单季收入还是个位数，那"合同负债往上"就只是交付慢的镜像。
- 01北方华创科技集团股份有限公司《2026年第三季度报告》2026-10-30 公告编号2026-060
- 02中微半导体设备（上海）股份有限公司《2026年第三季度报告》2026-10-31 巨潮ID 1225646685
- 03拓荆科技股份有限公司《2026年第三季度报告》2026-10-29 巨潮ID 1225636861
- 04盛美半导体设备（上海）股份有限公司《2026年第三季度报告》2026-10-29 巨潮ID 1225635956
- 05华海清科股份有限公司《2026年第三季度报告》2026-10-31 巨潮ID 1225648256
- 06沈阳芯源微电子设备股份有限公司《2026年第三季度报告》2026-10-30 巨潮ID 1225642763
- 07联合国商品贸易统计库 UN Comtrade 中国海关报送 HS848620 进口值与数量（2023、2025年度）
先把时点钉住。海关口径的10月数据，截至11月3日还没发布，按惯例要等到11月7日前后。所以"进口往下"现在是个预期，不是能核对的数。就算数出来，这个说法也只对一半：整机进口额从2022年186.93亿美元一路涨到2025年346.66亿，四年翻了近一倍，没有一年往下。
真正往下的是台数：15722台、11878台、11470台、10806台，四年少了31.3%。单价从118.9万美元涨到320.8万，翻了一倍七。
## 进口掉的不是金额，是台数和美日
按来源地拆开，2025年从美国进口20.7亿美元，比2024年的31.8亿少34.9%；从日本进口86.2亿，少10.5%，台数少了620台。两家合计从128.1亿掉到106.9亿，少16.6%。
同一时间，从新加坡进口57.0亿、涨17.1%，从马来西亚进口34.5亿、涨113.3%。这两个地方不是设备原产地。
| 来源地 | 2024 | 2025 | 同比% | 2025台数 |
|---|---|---|---|---|
| 荷兰 | 95.3 | 98.1 | 3.0 | 222 |
| 日本 | 96.3 | 86.2 | (10.5) | 4276 |
| 美国 | 31.8 | 20.7 | (34.9) | 1285 |
| 新加坡 | 48.7 | 57.0 | 17.1 | 1070 |
| 马来西亚 | 16.2 | 34.5 | 113.3 | 649 |
| 合计 | 335.1 | 346.7 | 3.5 | 10806 |
金额单位亿美元。荷兰那222台机器占了98.1亿，占全部进口金额的28%，台数只占2%。光刻机这一档没人替，也没人敢说替。
真正被挤掉的是日本那620台、美国那373台，对应到品类，是北方华创自己在三季报里点名"市占率稳步提升"的那几样：刻蚀、薄膜沉积、热处理、湿法清洗、离子注入、涂胶显影、键合。
```chart
type: group
title: 对美日进口在退，对东南亚在进（亿美元）
unit: 亿
labels: 2024, 2025
values: 128.1, 106.9
values2: 64.9, 91.5
legend: 美日|新马
```
所以"海外让出份额"得改成"海外在两个国家、特定档次上退出"。整个进口盘子是涨的，涨在东南亚那条通道上。
## 零件进口在涨，替代只发生在总装
半导体设备的零部件是另一个税号，HS848690。这条线2023年进口45.89亿美元、2024年49.12亿、2025年57.38亿，两年涨25.0%；同期整机进口涨26.5%。两条线几乎同速。
含义直接：整机在国产化，每台国产设备背后进口的零件没有减少。替代发生在总装那一层，价值量没跟着搬回国内。
整机在国产、零件靠进口，交付就卡在这一环。三季报那句"个别品类物料供应紧张"，说的是同一件事：整机厂在客户端争份额，回头在零件端还要求人。
界限：848690里含在华外资设备的维修备件，它只能证明零件进口没跟着往下走。
## 单季：合同负债横着走，存货往上堆
三季报单季收入113.06亿元，同比只增1.31%；合同负债53.01亿元，和二季度末的53.20亿基本没动。同比从47.04亿涨12.7%，可收入涨15.3%，蓄水池相对体量是缩的。
真正在动的是存货：一个季度增加33.93亿元，等于当季营业成本69.74亿的48.7%。折成天数，存货周转437天，去年同期408天。
钱从哪来？应付账款从去年三季度的116.28亿涨到180.19亿，涨54.9%，而收入只涨15.3%。周转天数从157天拉到233天，多占76天。
| 指标 | 2025Q3 | 2026Q3 | 变化 |
|---|---|---|---|
| 存货周转天数 | 408 | 438 | 30 |
| 应付账款周转天数 | 157 | 233 | 76 |
| 应收账款周转天数 | 73 | 72 | (1) |
| 净营业周期 | 324 | 277 | (47) |
| 收现/营业收入 | 90% | 111% | 21 |
净营业周期反而缩了47天——存货多占30天，上游多垫76天，一进一出，自己的现金周期短了。
单季净利润20.22亿只增5.18%，慢于收入；公司列的三条原因是研发投入增17.42%、销售费用增41.38%、毛利率降。
## 同行里只有它在掉毛利
六家摆一起看三季度单季毛利率，结论比"以价换量"难听：北方华创掉1.97个百分点，中微涨3.76个，拓荆涨8.05个。
如果"给大客户让价"是接份额必须付的入场费，那中微和拓荆在同一片市场里不该把毛利率抬起来。三家做的是同一件事，两家涨价、一家降价，说明那1.97个点不是行业的价，是它自己的价。
| 公司 | Q3收入同比% | 2025Q3毛利率% | 2026Q3毛利率% | 变动pp |
|---|---|---|---|---|
| 北方华创 | 1.3 | 40.31 | 38.33 | (1.97) |
| 中微公司 | 17.2 | 37.89 | 41.65 | 3.76 |
| 拓荆科技 | 1.8 | 34.42 | 42.47 | 8.05 |
| 盛美上海 | 7.6 | 47.48 | 43.95 | (3.53) |
| 华海清科 | 23.7 | 40.97 | 40.65 | (0.33) |
| 芯源微 | 55.6 | 30.12 | 22.38 | (7.74) |
芯源微是标准样本：收入涨55.6%，毛利率掉7.74个点。那才叫以价换量——量起来了，价让出去了，份额是买来的。
北方华创不一样。收入只涨1.3%，价又掉近2个点，量价都没赢。这不叫买份额，这叫单子接了、货交不出去、价格还得让。
```chart
type: bar
title: 三季度毛利率同比变动，同行两家在涨
unit: %
labels: 北方华创, 中微, 拓荆, 盛美, 华海清科, 芯源微
values: -1.97, 3.76, 8.05, -3.53, -0.33, -7.74
```
## 那55.9亿现金流不是客户多给的
前三季经营活动现金流净额55.91亿元，去年同期是负25.66亿，同比317.84%。公司自己写的解释是"营业收入增加，销售商品收到的货款增加"。
收现确实好了：销售商品收到现金349.13亿，去年同期245.49亿，收现/收入从90%抬到111%。但应收账款是平的，90.41亿对91.01亿。
差额在两头：购买商品支付的现金205.42亿，比去年同期少7.12亿；应付账款多62.2亿。存货增加52.7亿而采购付款减少，差额全部由供应商垫。
这55.9亿里，能算经营改善的只有收现恢复那一段，剩下是把账期往上游推。
## 账单最后落在谁头上
链条从终端往上：晶圆厂资本开支、设备厂新签单（先落在合同负债和存货，领先收入两到四个季度）、验收、确认收入、落到毛利率。海关数在第一环，三季报在第三第四环。
三季报原文写得很白：
> 2026年1-9月公司毛利率由41.41%下降到39.44%，主要原因为：一是多款新产品正处于客户端验证导入过程中，零部件迭代升级成本增加较多，二是公司为部分大批量采购客户提供了一定的商务优惠，以上均有助于进一步扩大市场份额。
这句话里三处让利全在：对客户让价，对自己让成本，对上游让账期。换回来的是同一件东西——份额。
利润往哪走？零件进口两年涨25%，整机进口涨26.5%，两条线同速。国产整机的"进口含量"没降，整机厂争来的每一块份额，都还要分一块给海外的零件供应商。利润留在最上游，不在总装。
所以"接住"这个说法，在量上成立，在钱上不成立。
## 最强的一条反驳
最强反驳不是数字错，是层级错：海关数是美元计价的设备进口，报表数是人民币计价的收入，中间隔着晶圆厂资本开支、验收周期、汇率，还有收入构成。
半年报只把收入拆成两项——电子工艺装备192.05亿元，占95.2%；电子元器件9.15亿元，占4.5%（2026年半年报，2026-08-26）。装备里集成电路、光伏、锂电各占多少，公司从没披露。
所以严格讲，那113亿单季收入里半导体设备占多少，我算不出来。拿这个数去对海关的半导体设备进口额，是拿体温计量血压。
第二个反驳更狠：进口降可能只是改道。新加坡涨17.1%、马来西亚涨113.3%，也可能不是转口，是中国晶圆厂把产能建到了那边——设备需求根本没回国内，"国产接住份额"就是另一种算法。这条我用现有数据排不掉，只能标出来。
我能守住的只有三点：美日合计少993台，同行两家毛利率在涨，应付账款多占76天。这三点不需要海关和报表对上账。
## 四个数字出来，这条判断就得改
一，10月海关数据，海关总署月度快讯，11月7日前后。848620美元金额同比若降超15%、且美日两项合计降超25%，"只是改道"就错了，是替代在加速。
二，北方华创2026年年报，2027年4月，看分产品毛利率。装备毛利率若回到41%以上，这1.97个点是新品导入的过路成本，不是定价权流失，整套解释作废。
三，2027年一季报，4月底，看资产负债表。合同负债若回到60亿以上、单季收入同比超20%，蓄水池重新蓄水、交付瓶颈解除，"接单不接钱"就说早了。
四，2026年全年HS848690进口额，2027年一季度海关年度数。若同比转负，"总装国产、零件进口"就说反了——零件也在替代，北方华创的毛利率该往上走，不是掉1.97个点。
来源：北方华创科技集团股份有限公司《2026年第三季度报告》2026-10-30 公告编号2026-060；《2026年半年度报告》2026-08-26；中微半导体设备（上海）股份有限公司《2026年第三季度报告》2026-10-31 巨潮ID 1225646685；拓荆科技股份有限公司《2026年第三季度报告》2026-10-29 巨潮ID 1225636861；盛美半导体设备（上海）股份有限公司《2026年第三季度报告》2026-10-29 巨潮ID 1225635956；华海清科股份有限公司《2026年第三季度报告》2026-10-31 巨潮ID 1225648256；沈阳芯源微电子设备股份有限公司《2026年第三季度报告》2026-10-30 巨潮ID 1225642763；联合国商品贸易统计库 UN Comtrade 中国海关报送 HS848620、HS848690 进口金额与数量（2023、2024、2025年度）。