德明利(001309)无公告跌7.64%、成交53.8亿，同日美光涨3.1%、西部数据涨7.3%、希捷涨4.3%，做模组…
- 三季度扣非37.53亿到47.53亿，上半年扣非100.47亿、一季度39.43亿，算下来二季度61.04亿，三季度环比降22%到39%
- 板块那天的情况：10月27日江波龙-2.11%、佰维-3.65%、朗科-4.51%、兆易创新-2.70%，德明利-7.64%，成交53.77亿，换手9.89%
- 它这九个月的存货从70.58亿涨到289.86亿，净增219.28亿，同期营业成本只有167.56亿，粗略算买进库的料约386亿，是卖掉的两倍多，池子里高价料占得越来越重
- 预告给的三季度区间是收入105.89亿到125.89亿、净利26.33亿到29.33亿
- 半年报披露，2026年上半年末存货跌价准备计提比例0.19%，2023年到2025年分别是1.34%、1.50%、0.50%
德明利(001309)无公告跌7.64%、成交53.8亿，同日美光涨3.1%、西部数据涨7.3%、希捷涨4.3%，做模组的和做颗粒的为什么反着走？
Demingli (001309) fell 7.64% without any announcement, with turnover of 5.38 billion yuan; on the same day Micron rose 3.1%, Western Digital rose 7.3%, and Seagate rose 4.3%. Why are module makers and chip makers moving in opposite directions?
The desk answers已核实Walnut, from public filings
这条的前提不成立：德明利不是“无公告”下跌。10月26日它刚披露2026年三季报（公告编号2026-093），再往前10月17日给过前三季度业绩预告（2026-092）。10月27日那7.64%，是对着这两份文件跌的。
但“做模组的和做颗粒的反着走”这件事是真的，也不是错杀：它们在同一条涨价链的两头，颗粒涨价对上游是收入，对模组厂是成本。
先把单季拆开。三季报只给累计数：前三季度收入285.55亿、营业成本167.56亿、归母净利87.46亿；上半年是169.11亿、89.42亿、60.17亿。相减得到三季度单季：收入116.45亿，成本78.15亿，净利27.29亿。单季毛利率32.89%，二季度38.84%，一季度57.42%。收入环比涨24.25%，净利环比只涨2.16%。收入一个台阶一个台阶往上走，毛利率一个台阶一个台阶往下走，这就是那天市场在卖的东西。
再看这份报表和业绩预告的关系。预告给的三季度区间是收入105.89亿到125.89亿、净利26.33亿到29.33亿。实际116.45亿、27.29亿——收入在区间里偏上，利润偏下。收入没问题，利润没跟上。
因为德明利买颗粒，不造颗粒。这条链从上到下是：云厂商买服务器，服务器里装企业级SSD和内存条，模组厂买颗粒把盘做出来，颗粒厂在晶圆上做电路。德明利在模组这一层，美光在颗粒这一层。半年报第18页写自己的生意：“根据方案需要采购存储晶圆，搭配自研主控、外采主控、PCB等材料或器件，通过封装、贴片、测试等工序后，形成闪存、内存存储模组，再将存储模组销售给下游渠道商、品牌商、行业客户、其他终端客户或消费者赚取利润。”晶圆是采购项，不是它的产品。
所以同一个价格上涨，在美光的报表里是售价，在德明利的报表里是成本。方向从一开始就是反的。
毛利率的数字能证明这一点。2025年上半年德明利综合毛利率5.03%，2026年上半年47.12%，半年报“毛利率比上年同期增减”一栏写着42.09个百分点。分产品更直接：嵌入式存储类从1.83%变成62.99%，固态硬盘类从4.29%变成39.90%，内存条类从8.55%变成38.47%。同一批人、同一条产线、同一类客户，一年时间毛利率差几十个点，这不是技术进步，是卖出去的货用的是涨价前买的料。
三季度的32.89%，就是这个时间差在收窄。半年报第125页写着存货“领用和发出时按月末一次加权平均法计价”——新买的高价颗粒一进库就把平均成本往上抬，卖一批、结转一批，成本就贵一批。它这九个月的存货从70.58亿涨到289.86亿，净增219.28亿，同期营业成本只有167.56亿，粗略算买进库的料约386亿，是卖掉的两倍多，池子里高价料占得越来越重。
上游那边正相反。美光9月30日发的2026财年四季报（截至9月3日），收入54.23亿美元，GAAP毛利率86.8%，上一季84.6%，再上一季74.4%。它的成本主要是晶圆厂折旧，售价涨多少，毛利就落多少。西数10月22日的2027财年一季报，收入同比增55%，GAAP毛利率57.8%，下季度收入指引同比增66%到72%。上游越紧，毛利率越陡；模组厂越贵买料，毛利率越薄。
第二层原因在资产负债表上。三季报：存货289.86亿，年初70.58亿，增310.66%，公司给的理由是“报告期内公司企业级存储业务规模及在手订单量快速增长，存货规模增加主要系公司为企业级存储业务备货所致”。短期借款101.50亿，年初34.72亿；长期借款91.00亿，年初9.52亿；加一年内到期的10.61亿，有息负债203.11亿。账上货币资金9.51亿。净欠193.6亿，股东权益120.27亿，资产负债率70.10%。前三季度经营活动现金流净额-136.91亿，同期净利润87.46亿。利润是账面的，垫进去的是现金。
这289.86亿还有一层脆弱性。半年报披露，2026年上半年末存货跌价准备计提比例0.19%，2023年到2025年分别是1.34%、1.50%、0.50%。基本按成本挂着。价格再涨，它是资产；价格一转向，它就是减值的来源，而且计提比例越低，转向时的落差越大。
同行能对上。江波龙10月15日的业绩预告：三季度收入139亿到159亿，二季度141.79亿，环比基本没动；三季度扣非37.53亿到47.53亿，上半年扣非100.47亿、一季度39.43亿，算下来二季度61.04亿，三季度环比降22%到39%。它自己写的理由：“受前期较高价格基数影响，DRAM和NAND Flash价格涨幅较前期明显收窄，存储价格上涨对公司盈利的边际贡献显著趋缓。”这就是模组厂三季度的通用剧本，德明利只是没在这份三季报里写这句话。
江波龙还写了两句值得对照的：“实现第三季度经营性现金流量净额转正的重大跨越”“实现期末资产负债率低于50%”。同样的行业、同样的涨价，一家现金流转正、负债率往下走，一家现金流-136.91亿、负债率70.10%。德明利的问题不只是站在模组这个位置，还在于它选的扩张方式。
佰维的三季报还没出，10月24日只有一份生产经营情况公告，讲的是行业景气、回购和分红，没有三季度的数。所以现在拿不出第三家模组厂的单季毛利率。谁要说“模组厂集体承压”，手上目前只有德明利一份正式报表加江波龙一份预告。
还有一处要拆掉：西数和希捷不是颗粒厂，是硬盘厂。卖的是磁头、碟片、盘片，这轮涨的理由是AI数据中心近线硬盘的交期和价格，跟NAND颗粒不是一条链。把它们和德明利摆在一起，标签都是“存储”，生意差着两层。另外那三个涨幅是10月27日美股盘中的数，北京时间已经是28日凌晨，A股当天下午就收完了——两个价格不在同一个时点上。
板块那天的情况：10月27日江波龙-2.11%、佰维-3.65%、朗科-4.51%、兆易创新-2.70%，德明利-7.64%，成交53.77亿，换手9.89%。10月17日预告出来后，它10月19日到23日五个交易日涨了21.84%，10月26日、27日两天又跌掉10.76%。多跌出来的那三五个点，只能由它自己的报表解释，推不到板块头上。
反方最强的一击是这个：32.89%仍然是2025年全年14.81%的两倍多，27.29亿是历史第二高的单季净利；289.86亿存货如果四季度价格再涨一轮，会重新把毛利率推上去；两天跌10.76%更像预告后那21.84%涨幅的获利盘在走，不是基本面转折。
这话有成立的部分，但有一个关键环节站不住。如果毛利率下滑真是“企业级占比上升、结构变了”，数字不支持：上半年企业级收入48.15亿，占169.11亿的28.47%；前三季度82.37亿，占285.55亿的28.85%，只动了0.38个百分点；倒推三季度单季企业级34.22亿，占116.45亿的29.39%。三个季度几乎是一条平线，解释不了5.95个点的降幅。而且企业级到现在没有单独披露过毛利率，“企业级毛利低”是没有任何文件支持的说法，要等年报的分产品表。
什么情况下上面这套判断算错：四季度单季毛利率回到37%以上，说明三季度是产品结构或者会计节奏的问题，不是便宜料卖完了；或者四季度存货环比下降、经营现金流转正、毛利率不再往下走，说明这轮备货已经转成周转；再或者TrendForce公布的四季度NAND合约价环比涨幅重新扩大到二季度的水平，模组厂手里这289.86亿存货会先受益一轮，我这条链的顺序就讲反了。
反过来，如果四季度毛利率跌破30%、存货还往300亿以上走、短期借款再上一个台阶，那就是同一个方向走得更远，不是修正。
来源：深圳市德明利技术股份有限公司《2026年第三季度报告》，2026年10月26日，公告编号2026-093；《2026年前三季度业绩预告》，2026年10月17日，公告编号2026-092；《2026年半年度报告》，2026年8月28日，正文未标公告编号；深圳市江波龙电子股份有限公司《2026年前三季度业绩预告》，2026年10月15日，公告编号2026-117；深圳佰维存储科技股份有限公司《关于公司近期生产经营情况的公告》，2026年10月24日，公告编号2026-100；Micron Technology, Inc.《Micron Technology, Inc. Reports Record Fiscal Fourth-Quarter and Full-Year 2026 Results》，2026年9月30日，SEC 0000723125-26-000018；Western Digital Corporation《WD Reports Fiscal First Quarter 2027 Financial Results》，2026年10月22日，SEC 0001628280-26-066882。行情为腾讯财经2026年10月27日收盘数据。
The premise of this piece does not hold: Demingli's decline was not "without announcements." On October 26 it had just disclosed its 2026 third-quarter report (announcement no. 2026-093), and before that, on October 17, it had issued a pre-announcement of results for the first three quarters (2026-092). The 7.64% drop on October 27 was a reaction to these two documents.
But the claim that "module makers and wafer makers move in opposite directions" is true, and it is not a mistaken selloff: they sit at the two ends of the same price-increase chain. For upstream, higher wafer prices mean revenue; for module makers, they mean cost.
Let's break out the single quarter first. The third-quarter report gives only cumulative figures: first-three-quarter revenue 285.55 yi yuan, operating cost 167.56 yi yuan, net profit attributable to parent 87.46 yi yuan; first half was 169.11 yi yuan, 89.42 yi yuan, 60.17 yi yuan. Subtracting gives the third-quarter single quarter: revenue 116.45 yi yuan, cost 78.15 yi yuan, net profit 27.29 yi yuan. Single-quarter gross margin 32.89%, second quarter 38.84%, first quarter 57.42%. Revenue rose 24.25% quarter-on-quarter, while net profit rose only 2.16% quarter-on-quarter. Revenue climbing step by step while gross margin falls step by step—that is what the market was selling that day.
Now look at the relationship between this report and the pre-announcement. The pre-announced third-quarter range was revenue 105.89 yi yuan to 125.89 yi yuan and net profit 26.33 yi yuan to 29.33 yi yuan. Actual was 116.45 yi yuan and 27.29 yi yuan—revenue toward the upper part of the range, profit toward the lower part. Revenue was fine; profit did not keep up.
Because Demingli buys wafers, it does not make them. The chain from top to bottom is: cloud providers buy servers, servers are fitted with enterprise SSDs and memory modules, module makers buy wafers to build the drives, and wafer makers build circuits on wafers. Demingli sits at the module layer; Micron sits at the wafer layer. Page 18 of the interim report describes its business: "purchase memory wafers as required by the solution, pair them with self-developed controllers, externally sourced controllers, PCBs and other materials or components, and after packaging, SMT, testing and other processes, form flash memory and memory storage modules, then sell the storage modules to downstream channel distributors, brand owners, industry customers, other end customers or consumers to earn profit." Wafers are a purchased item, not its product.
So the same price increase is a selling price in Micron's report and a cost in Demingli's report. The direction is opposite from the very beginning.
The gross margin numbers prove this. In the first half of 2025 Demingli's consolidated gross margin was 5.03%; in the first half of 2026 it was 47.12%, and the interim report's "increase/decrease in gross margin versus the same period last year" line shows 42.09 percentage points. By product it is even more direct: embedded storage went from 1.83% to 62.99%, solid-state drives from 4.29% to 39.90%, and memory modules from 8.55% to 38.47%. Same people, same production line, same type of customers, yet gross margin differs by dozens of points in one year—this is not technological progress, it is that the goods sold used materials bought before the price increase.
The third quarter's 32.89% is this time lag narrowing. Page 125 of the interim report says inventory is "valued using the monthly weighted average method when requisitioned and issued"—newly purchased high-priced wafers raise the average cost as soon as they enter the warehouse, and each batch sold and carried over makes the cost of the next batch more expensive. Over these nine months its inventory rose from 70.58 yi yuan to 289.86 yi yuan, a net increase of 219.28 yi yuan, while operating cost over the same period was only 167.56 yi yuan. Roughly calculated, materials brought into the warehouse were about 386 yi yuan, more than twice what was sold, and high-priced material makes up an increasing share of the pool.
Upstream is the opposite. Micron's fiscal 2026 fourth-quarter report released on September 30 (quarter ended September 3) showed revenue of 54.23 US dollars yi, GAAP gross margin 86.8%, versus 84.6% the previous quarter and 74.4% the quarter before that. Its cost is mainly fab depreciation, so however much the selling price rises, that much falls to gross profit. Western Digital's fiscal 2027 first-quarter report on October 22 showed revenue up 55% year-on-year, GAAP gross margin 57.8%, and next-quarter revenue guidance up 66% to 72% year-on-year. The tighter upstream is, the steeper the gross margin; the more expensive the materials module makers buy, the thinner their gross margin.
The second layer of cause is on the balance sheet. Third-quarter report: inventory 289.86 yi yuan, versus 70.58 yi yuan at the start of the year, up 310.66%; the company's stated reason is "during the reporting period the company's enterprise storage business scale and order backlog grew rapidly, and the increase in inventory scale is mainly due to the company stocking up for its enterprise storage business." Short-term borrowings 101.50 yi yuan, versus 34.72 yi yuan at the start of the year; long-term borrowings 91.00 yi yuan, versus 9.52 yi yuan at the start of the year; adding 10.61 yi yuan due within one year, interest-bearing debt is 203.11 yi yuan. Cash on the books is 9.51 yi yuan. Net debt is 193.6 yi yuan, shareholders' equity is 120.27 yi yuan, and the debt-to-asset ratio is 70.10%. Net cash flow from operating activities in the first three quarters was -136.91 yi yuan, while net profit over the same period was 87.46 yi yuan. Profit is on the books; what was advanced is cash.
This 289.86 yi yuan has another layer of fragility. The interim report discloses that at the end of the first half of 2026 the inventory write-down provision ratio was 0.19%, versus 1.34%, 1.50% and 0.50% in 2023 to 2025 respectively. It is basically carried at cost. If prices rise further, it is an asset; if prices turn, it becomes a source of impairment, and the lower the provision ratio, the bigger the drop when it turns.
Peers can be matched up. Jiangbolong's October 15 earnings forecast: third-quarter revenue 13.9 billion to 15.9 billion, second quarter 14.179 billion, basically flat quarter-on-quarter; third-quarter excluding non-recurring items 3.753 billion to 4.753 billion, first half excluding non-recurring items 10.047 billion, first quarter 3.943 billion, which works out to 6.104 billion in the second quarter, third quarter down 22% to 39% quarter-on-quarter. Its own stated reason: "Affected by the higher price base in the earlier period, the price increases of DRAM and NAND Flash narrowed significantly compared with the earlier period, and the marginal contribution of memory price increases to the company's profitability slowed markedly." This is the generic script for module makers in the third quarter; Demingli simply did not write this sentence in this third-quarter report.
Jiangbolong also wrote two sentences worth comparing: "achieved a major leap in turning the third-quarter net operating cash flow positive" and "achieved a debt-to-asset ratio below 50% at period end." Same industry, same price increases, one company's cash flow turned positive and its debt ratio moved down, the other's cash flow was -13.691 billion and its debt ratio 70.10%. Demingli's problem is not just standing in the module position, but also the expansion method it chose.
Baiwei's third-quarter report is not out yet; on October 24 there was only a production and operations announcement, discussing industry prosperity, buybacks and dividends, with no third-quarter numbers. So right now a third module maker's single-quarter gross margin cannot be produced. Anyone who wants to say "module makers are under collective pressure" currently has only one formal report from Demingli plus one forecast from Jiangbolong in hand.
There is another point to dismantle: Western Digital and Seagate are not chip makers, they are hard drive makers. What they sell is heads, platters, and disks; the reason for this round of increases is the lead time and prices of nearline hard drives for AI data centers, which is not the same chain as NAND chips. Placing them together with Demingli, the label is all "storage," but the businesses are two layers apart. In addition, those three gains are intraday US stock figures from October 27; Beijing time was already the early hours of the 28th, and the A-share market had already closed that afternoon—the two prices are not at the same point in time.
The sector's situation that day: on October 27, Jiangbolong -2.11%, Baiwei -3.65%, Langke -4.51%, GigaDevice -2.70%, Demingli -7.64%, turnover 5.377 billion, turnover rate 9.89%. After the forecast came out on October 17, it rose 21.84% over the five trading days from October 19 to 23, then fell 10.76% over the two days of October 26 and 27. The extra three to five points of decline can only be explained by its own report and cannot be pushed onto the sector.
The strongest counterargument is this: 32.89% is still more than double the full-year 2025 figure of 14.81%, 2.729 billion is the second-highest single-quarter net profit in history; if prices rise another round in the fourth quarter, the 28.986 billion of inventory will push gross margin back up; the two-day 10.76% drop looks more like profit-taking after the 21.84% gain following the forecast, not a fundamental turning point.
Part of this holds, but one key link does not stand up. If the gross margin decline really were "enterprise share rising, structure changed," the numbers do not support it: first-half enterprise revenue 4.815 billion, 28.47% of 16.911 billion; first three quarters 8.237 billion, 28.85% of 28.555 billion, moving only 0.38 percentage points; back-calculating third-quarter single-quarter enterprise 3.422 billion, 29.39% of 11.645 billion. The three quarters are almost a flat line and cannot explain a 5.95-point decline. Moreover, enterprise gross margin has never been separately disclosed so far; "enterprise gross margin is low" is a statement with no document supporting it, and we must wait for the annual report's product breakdown table.
Under what circumstances would the above judgment be wrong: if fourth-quarter single-quarter gross margin returns above 37%, it would show that the third quarter was a matter of product mix or accounting timing, not that cheap materials had been sold out; or if fourth-quarter inventory declines quarter-on-quarter, operating cash flow turns positive, and gross margin stops falling, it would show that this round of stockpiling has turned into turnover; or if TrendForce's published fourth-quarter NAND contract price quarter-on-quarter increase re-expands to the second-quarter level, the 28.986 billion of inventory in module makers' hands would benefit first, and the order of my chain would be reversed.
Conversely, if fourth-quarter gross margin falls below 30%, inventory still moves above 30 billion, and short-term borrowings step up another level, then it is the same direction going further, not a correction.
Sources: Shenzhen Deminglin Technology Co., Ltd. "2026 Third Quarter Report," October 26, 2026, announcement no. 2026-093; "2026 First Three Quarters Earnings Forecast," October 17, 2026, announcement no. 2026-092; "2026 Semi-Annual Report," August 28, 2026, main text does not indicate an announcement number; Shenzhen Jiangbolong Electronics Co., Ltd. "2026 First Three Quarters Earnings Forecast," October 15, 2026, announcement no. 2026-117; Shenzhen Baiwei Storage Technology Co., Ltd. "Announcement on the Company's Recent Production and Operations," October 24, 2026, announcement no. 2026-100; Micron Technology, Inc. "Micron Technology, Inc. Reports Record Fiscal Fourth-Quarter and Full-Year 2026 Results," September 30, 2026, SEC 0000723125-26-000018; Western Digital Corporation "WD Reports Fiscal First Quarter 2027 Financial Results," October 22, 2026, SEC 0001628280-26-066882. Market data are Tencent Finance closing data for October 27, 2026.