On the day of its earnings report, Junzheng Co., Ltd. (300223) rose 18.37% with a turnover of 72.7 yi, while GigaDevice, which also makes memory chips, rose 7.77% without any announcement. Was this 10.6 percentage point gap given by the earnings report or lifted by the sector?
- Biwin fell from 55.98% to 42.81%, and Demingli fell from 38.84% to 32.89%
- GigaDevice's 7.77% must be viewed separately.
- What would count as being wrong Fourth-quarter single-quarter gross margin below 65.93%.
- Accounts receivable 12.32 yi yuan, up 164% from the start of the year
- When the memory chain rises 3.32% more than ChiNext, it usually rises another 2.90 points
On the day of its earnings report, Junzheng Co., Ltd. (300223) rose 18.37% with a turnover of 72.7 yi, while GigaDevice, which also makes memory chips, rose 7.77% without any announcement. Was this 10.6 percentage point gap given by the earnings report or lifted by the sector?
君正股份(300223)财报当天涨18.37%、成交72.7亿，而同做存储芯片的兆易创新没有任何公告也涨7.77%，这10.6个百分点的差距是财报给的还是板块抬的？
The desk answers已核实Walnut, from public filings
These 18.37 percentage points are not an either-or choice; they are three layers stacked together: the ChiNext Index's rebound that day contributed 4.4 points, the portion by which the memory chain rose more than the broad market contributed 2.9 points, and Junzheng Co., Ltd. (300223)'s own financial report contributed 10.7 points. 18.37 minus 7.77 equals 10.60; this number is almost the same as 10.7, and it is not a coincidence — in GigaDevice's 7.77%, not a single cent came from the financial report.
On October 28, the ChiNext Index rose 3.40%, after having just fallen 3.41% the previous day; this was a whole-market matter. On the same day, the median gain of seven companies in the memory chain was 6.69%: Puya 11.38%, Biwin 8.36%, GigaDevice 7.77%, Hengshuo 6.69%, Dongxin 5.04%, Longsys 4.87%, Demingli 2.93%. The entire chain rose together, indicating that one layer that day was sector-wide and unrelated to Junzheng itself.
Based on the actual relationship between Junzheng and these two sets of numbers over the past 100 trading days: when the ChiNext Index rises 3.40%, it usually rises 4.43 points in tandem; when the memory chain rises 3.32% more than ChiNext, it usually rises another 2.90 points; plus the 0.39-point constant by which it has long outperformed the sector, the three items together come to 7.72 points. Subtracting 7.72 from 18.37 leaves 10.66 points that were extra on the financial report day and had not appeared in four months — over the past 100 days, its average deviation from this line was only 2.93 points; on October 28 it deviated by 10.66 points, three and a half times.
These 10.66 points have a clear timestamp. That day it opened directly with a gap up to 105.24 yuan, 11.25% above the previous close, while the median opening gain of peers was only 3.73%; the trade that read the filing overnight did most of the work. During the session it moved from 105.24 to 111.98, rising another 6.40%, while the median intraday gain of peers was only 2.16%; this stretch was capital接力, not new news. Full-day turnover was 72.75 yi yuan, 3.5 times the 21.01 yi yuan average of the previous five trading days, with turnover of 15.71%.
The 11 points given by the financial report land on one line item
In the third quarter alone, operating revenue was 31.45 yi yuan, up 31.1% from 23.97 yi yuan in the second quarter and up 164.85% year-on-year. Net profit attributable to the parent was 14.19 yi yuan, up 61.5% from 8.79 yi yuan in the second quarter and up 2595.49% year-on-year.
The key is not these two numbers, but the speed gap between cost and revenue: single-quarter operating cost was 10.71 yi yuan, only 4.2% more than 10.28 yi yuan in the second quarter. Revenue rose three tenths, cost rose 4%, gross profit increased from 13.70 yi yuan to 20.73 yi yuan, up 51.4%, and gross margin rose from 57.12% to 65.93%, adding 8.8 points in one quarter.
This profit was not saved out or obtained by selling assets. Single-quarter net profit attributable to the parent excluding non-recurring items was 14.29 yi yuan, slightly higher than the 14.19 yi yuan attributable to the parent; non-recurring gains and losses were negative 0.10 yi yuan, and government subsidies were only 175 wan yuan. The price increase all sank into the "operating cost" line: revenue was settled at post-increase prices, while foundry prices on the cost side rose only a small portion.
The company's DRAM chip market demand is strong, and tight market supply has led to continuous product price increases; Flash chip market demand continues to grow, especially in emerging fields such as AI servers...
This is the original wording from the section "Circumstances and reasons for changes in major accounting data and financial indicators" in the third-quarter report. The quarterly report only discusses direction; to see how volume and price split, one must look at the research record posted the same day:
In the first three quarters, DRAM shipment volume increased by about 20% year-on-year... DRAM revenue growth mainly came from ASP increases; Nor Flash shipment volume grew substantially... mainly driven by volume.
DRAM shipments rose only two tenths, while total memory chip revenue was 49.16 yi yuan, up 130.73% — the difference in between is price, not volume. In the product mix, DRAM accounts for more than five tenths, Flash more than three tenths, and SRAM less than one tenth. At this point, the nature of those 11 points is settled: the market was buying "price increases are still continuing," not "more was sold."
On the same chain, the other half of companies are losing gross margin
| Company | October 28 change | 2026Q2 single-quarter gross margin | 2026Q3 single-quarter gross margin | Q3 report disclosure |
|---|---|---|---|---|
| 君正股份 | +18.37% | 57.12% | 65.93% | October 28 |
| 东芯股份 | +5.04% | 74.49% | 79.31% | October 27 |
| 兆易创新 | +7.77% | 66.57% | Not disclosed | Not disclosed |
| 江波龙 | +4.87% | 61.26% | Not disclosed | Not disclosed |
| 普冉股份 | +11.38% | 54.87% | Not disclosed | Not disclosed |
| 佰维存储 | +8.36% | 55.98% | 42.81% | October 29 |
| 德明利 | +2.93% | 38.84% | 32.89% | October 26 |
There is a directional divide in the table: chip designers (Junzheng, Dongxin) saw single-quarter gross margin expand, while those buying wafers to make modules (Biwin, Demingli) saw it contract. Biwin fell from 55.98% to 42.81%, and Demingli fell from 38.84% to 32.89%; over the same period, Demingli's single-quarter revenue even rose from 93.72 yi yuan to 116.45 yi yuan. Revenue up and gross margin down means wafer price increases outpaced module selling prices.
Junzheng itself explained this relationship clearly in the research: foundry prices are indeed rising, but "the price increase of pure foundries is relatively limited," and "the overall increase in foundry costs has been passed on to customers through product pricing adjustments." In the same price increase, whoever holds bargaining power sees the statements tilt their way.
GigaDevice's 7.77% must be viewed separately. As of October 28, it had not yet disclosed its third-quarter report; that day there was only a monthly repurchase progress announcement, which contained no operating data. So its 7.77% was purely sector-driven — the market was buying the expectation that "its third quarter will not be bad either," because in the first half its gross margin had already moved from 37.21% in the same period last year to 63.13%.
Looking at Dongxin in reverse: it disclosed its third-quarter report on October 27, with a single-quarter gross margin of 79.31%, higher than Junzheng's, yet it fell 6.94% that day and rose 5.04% the next day, a combined negative 2.25% over the two days. Also a design-side company, also expanding gross margin, yet it did not earn an 11-point excess return.
There is also a stretch of oversold
In the 20 trading days before October 28, Junzheng fell from 146.10 yuan on September 21 to 94.60 yuan, down 35.2%, after hitting an intraday high of 161.55 yuan on August 18. The closing price of 111.98 yuan was still a stretch below where it was a month earlier. The company itself admitted in its October 27 research session that "this year the stock price has indeed been affected by macroeconomic factors and the global sector, and the secondary market has come under relatively heavy pressure."
For a stock that has fallen 35%, running into a report with single-quarter profit up 61% quarter-on-quarter, paired with a line that "fourth-quarter gross margin will continue to grow," 18 points is not outrageous. But it is oversold plus better-than-expected, not a trend breakout. These two things will diverge sharply in the subsequent price action.
How a smart person would rebut
The strongest rebuttal is: these 11 points are buying price, not volume, and price is the least sustainable.
The company wrote the answer in its research record — next year's memory shipments "will maintain a level similar to this year," 2027 DRAM growth "may mainly still come from price increases," and new capacity "is expected to come on stream only gradually after 2027." Over the next four quarters, the slope of profit depends entirely on whether price increases can continue. Asked how computing chips would develop going forward, the chairman's line was even more blunt: "The performance boost brought by shortage-driven price increases is temporary."
Where does the 65.93% gross margin sit in Junzheng's own history: single-quarter 31.63% in the third quarter of 2025, and 34.09% for full-year 2025. In one year it lifted single-quarter gross margin by 34.3 percentage points — this is the shape of a cycle top, not the shape of a new normal.
There are also things in the statements worth watching. Accounts receivable 12.32 yi yuan, up 164% from the start of the year; inventory 35.99 yi yuan, up 21% from the start of the year; asset impairment loss 0.95 yi yuan, up 68% year-on-year — still accruing inventory write-downs during a price-increase cycle shows that not all product lines are making money. Operating cash flow 18.96 yi yuan against net profit attributable to parent of 26.16 yi yuan, or 72.5%. Annualizing the third-quarter single-quarter 14.19 yi yuan, the 473 yi yuan market cap corresponds to about 8.3x, and the market clearly is not treating third-quarter earnings as normal.
What would count as being wrong
- Fourth-quarter single-quarter gross margin below 65.93%. The company said it "expects memory gross margin to continue growing in the fourth quarter"; if it fails to deliver that line, it is void, and the pricing basis for those 11 points is void with it.
- If Zhaoyi's upcoming third-quarter report shows no quarter-on-quarter gross margin expansion, that means the price increases landed only in Junzheng's own customer structure, not industry-wide, and the 2.9-point sector premium will be given back.
- If first-quarter 2027 revenue does not grow quarter-on-quarter and inventory keeps climbing, that is channel stockpiling, and what follows is price declines.
Sources: Beijing Junzheng Integrated Circuit Co., Ltd. "2026 Third Quarter Report," October 28, 2026, announcement no. 2026-077; "Announcement of Resolutions of the 15th Meeting of the Sixth Board of Directors," October 28, 2026, announcement no. 2026-076; "Investor Relations Activity Record Form," October 27, 2026, no. 20261027; Dongxin Semiconductor "2026 Third Quarter Report," October 27, 2026, Cninfo announcement 1225626412; Demingli "2026 Third Quarter Report," October 26, 2026, Cninfo announcement 1225620615; Biwin Storage "2026 Third Quarter Report," October 29, 2026, Cninfo announcement 1225638424; Zhaoyi Innovation "Progress Announcement on Repurchase of A-Shares by Centralized Bidding," October 28, 2026, Cninfo announcement 1225630578; Zhaoyi Innovation, Longsys, and Puya Semiconductor Cninfo announcement lists (as of October 28, 2026, no third-quarter reports seen); single-quarter data for each company calculated by subtracting cumulative operating revenue and operating cost figures in periodic reports; price changes and turnover taken from exchange daily lines.
这18.37个百分点不是二选一，是三层叠起来的：创业板指当天的反弹贡献4.4个点，存储链比大盘多涨的那一截贡献2.9个点，君正股份（300223）自己的财报贡献10.7个点。18.37减7.77等于10.60，这个数与10.7几乎一样，不是巧合——兆易创新那7.77%里，一分钱财报都没有。
10月28日创业板指涨3.40%，前一天刚跌3.41%，这是全市场的事。同一天存储链七家公司涨幅中位数6.69%：普冉11.38%、佰维8.36%、兆易7.77%、恒烁6.69%、东芯5.04%、江波龙4.87%、德明利2.93%。整条链一起涨，说明那天有一层是板块性的，跟君正自己无关。
按过去100个交易日君正跟这两组数字的实际关系算：创业板指涨3.40%，它通常跟涨4.43个点；存储链比创业板多涨3.32%，它通常再多涨2.90个点；再加上它长期跑在板块前面那0.39个点的常数，三项合起来7.72个点。18.37减去7.72，剩下的10.66个点是财报当天多出来的，四个月里没出现过——过去100天它偏离这条线的平均幅度只有2.93个点，10月28日偏了10.66个点，三倍半。
这10.66个点有明确的时间戳。那天开盘直接跳空到105.24元，比前收盘高11.25%，而同行开盘的中位数只高3.73%；隔夜读完文件的那一笔，就把大头打完了。盘中它从105.24走到111.98，又涨6.40%，同行盘中中位数只有2.16%，这一段是资金接力，不是新消息。全天成交72.75亿元，是前五个交易日均值21.01亿元的3.5倍，换手15.71%。
财报给的11个点，落在一行科目上
三季度单季，营业收入31.45亿元，比二季度的23.97亿元多31.1%，比去年同期多164.85%。归母净利润14.19亿元，比二季度的8.79亿元多61.5%，比去年同期多2595.49%。
关键不是这两个数，是成本和收入的速度差：单季营业成本10.71亿元，比二季度的10.28亿元只多4.2%。收入涨三成，成本涨4%，毛利额从13.70亿元变成20.73亿元，多出51.4%，毛利率从57.12%抬到65.93%，一个季度加了8.8个点。
这笔利润不是省出来的，也不是卖资产来的。单季扣非归母净利润14.29亿元，比归母的14.19亿元还高一点，非经常性损益是负的0.10亿元，政府补助只有175万元。涨价全部沉在"营业成本"这一行：收入按涨价后的价格结算，成本端的代工价只涨了一小部分。
公司DRAM芯片市场需求旺盛，市场供应紧张导致产品持续涨价，Flash芯片市场需求不断增长，尤其在AI服务器等新兴领域……
这是三季报里"主要会计数据和财务指标发生变动的情况及原因"那一段的原话。季报只讲方向，量价怎么分，要看同一天挂出来的那份调研记录：
前三季度DRAM出货量同比增幅约20%上下……DRAM收入增长主要来自ASP提升；Nor Flash出货量大幅增长……主要还是量的提升。
DRAM出货只增了两成，存储芯片整体收入49.16亿元、涨了130.73%——这中间差出来的部分是价格，不是量。产品结构里DRAM占五成多、Flash占三成多、SRAM不到一成。到这一步，那11个点的性质就定了：市场买的是"涨价还在继续"，不是"卖得更多了"。
同一条链上，另一半公司在亏毛利
| 公司 | 10月28日涨跌 | 2026Q2单季毛利率 | 2026Q3单季毛利率 | 三季报披露 |
|---|---|---|---|---|
| 君正股份 | +18.37% | 57.12% | 65.93% | 10月28日 |
| 东芯股份 | +5.04% | 74.49% | 79.31% | 10月27日 |
| 兆易创新 | +7.77% | 66.57% | 未披露 | 未披露 |
| 江波龙 | +4.87% | 61.26% | 未披露 | 未披露 |
| 普冉股份 | +11.38% | 54.87% | 未披露 | 未披露 |
| 佰维存储 | +8.36% | 55.98% | 42.81% | 10月29日 |
| 德明利 | +2.93% | 38.84% | 32.89% | 10月26日 |
表里有个方向性的分野：做芯片设计的（君正、东芯）单季毛利率在扩，买晶圆做模组的（佰维、德明利）在缩。佰维从55.98%掉到42.81%，德明利从38.84%掉到32.89%，同期德明利单季收入还从93.72亿元涨到116.45亿元。收入涨、毛利率掉，说明晶圆涨价快过模组售价。
君正自己在调研里把这层关系说透了：代工价确实在涨，但"纯代工厂价格涨幅相对有限"，而且"总体代工成本上升已通过产品定价调整转嫁给客户"。同一场涨价，议价权在谁手上，报表就往谁那边偏。
兆易创新那7.77%要单独看。截至10月28日它还没披露三季报，那天只有一份每月都发的回购进展公告，里面没有经营数据。所以它的7.77%是纯板块——市场买的是"它三季度也不会差"的预期，因为上半年它的毛利率已经从去年同期的37.21%走到63.13%。
反过来看东芯：10月27日披露三季报，单季毛利率79.31%比君正还高，可当天跌6.94%、第二天涨5.04%，两天合起来是负2.25%。同样是设计端、同样在扩毛利，没换来11个点的超额。
还有一段是超跌
10月28日之前的20个交易日里，君正从9月21日的146.10元跌到94.60元，跌了35.2%，8月18日盘中最高到过161.55元。111.98元的收盘价，离一个月前还差一截。公司自己在10月27日的调研里承认"今年股价确实受到宏观因素及全球板块的影响，二级市场承压较重"。
一个跌了35%的股票，撞上单季利润环比增61%的财报，再配一句"四季度毛利率继续增长"，18个点不算离谱。但它是超跌加超预期，不是趋势突破。这两件事在后面的走势里会分得很开。
一个聪明人会怎么反驳
最强的反驳是：这11个点买的是价，不是量，而价最不可持续。
公司自己把答案写在调研记录里了——明年存储出货量"将维持与今年相近水平"，2027年DRAM增长"可能主要还是来自价格提升"，新增产能"预计2027年后才能陆续到位"。未来四个季度，利润的斜率完全取决于涨价能不能延续。被问到计算芯片往后怎么走时，董事长那句话更直白："因缺货涨价带来的业绩提升是暂时的。"
65.93%的毛利率放在君正自己的历史里是什么位置：2025年三季度单季31.63%，2025年全年34.09%。一年时间把单季毛利率抬了34.3个百分点，这是周期顶部的形状，不是新常态的形状。
报表里也有该盯的地方。应收账款12.32亿元，比年初增164%；存货35.99亿元，比年初增21%；资产减值损失0.95亿元，比去年同期增68%——涨价周期里还在计提存货跌价，说明不是所有产品线都在赚钱。经营现金流18.96亿元对归母净利26.16亿元，是72.5%。按三季度单季14.19亿元年化，473亿元市值对应约8.3倍，市场显然没把三季度的盈利当成常态。
什么情况算看错
- 四季度单季毛利率低于65.93%。公司说"预计四季度存储毛利率会继续增长"，做不到这句话就废了，那11个点的定价基础也跟着废。
- 兆易接下来披露的三季报，如果它单季毛利率环比没扩，说明涨价只落在君正自己的客户结构里，不是行业性的，那2.9个点的板块溢价会还回去。
- 2027年一季度收入环比不增、存货继续往上，就是渠道在囤货，之后跟着来的是跌价。
来源：北京君正集成电路股份有限公司《2026年第三季度报告》，2026年10月28日，公告编号2026-077；《第六届董事会第十五次会议决议公告》，2026年10月28日，公告编号2026-076；《投资者关系活动记录表》，2026年10月27日，编号20261027；东芯半导体《2026年第三季度报告》，2026年10月27日，巨潮公告1225626412；德明利《2026年三季度报告》，2026年10月26日，巨潮公告1225620615；佰维存储《2026年第三季度报告》，2026年10月29日，巨潮公告1225638424；兆易创新《关于以集中竞价交易方式回购A股股份的进展公告》，2026年10月28日，巨潮公告1225630578；兆易创新、江波龙、普冉股份巨潮公告列表（截至2026年10月28日，未见三季报）；各公司单季数据由定期报告营业收入、营业成本累计数相减算得，涨跌幅与成交额取交易所日线。
## 先排除一个解释：这是不是A股资金的事
君正8月25日在香港上市（03223.HK）。10月28日同一天，港股君正涨18.94%、港股兆易涨9.91%，差9.03个点；A股这边18.37%和7.77%，差10.60个点。两个市场、两套交易制度、两种投资者结构，量出几乎一样的缺口。
```chart
type: group
title: 同一天同一批公司，港股全部跑在A股前面
unit: %
labels: 君正股份, 兆易创新, 华虹半导体, 中芯国际
values: 18.37, 7.77, 3.62, 2.00
values2: 18.94, 9.91, 8.44, 5.43
legend: A股|港股
```
那天恒生指数只涨0.23%。港股没有创业板指，它照样涨了18.94%，所以"创业板指反弹给君正贡献了几个点"这个算法在这里不成立——真有那么大的指数贡献，港股就该少涨，而它多涨了。
尺子不止一对。华虹半导体港股8.44%、A股3.62%；中芯国际港股5.43%、A股2.00%。四对配对，港股全部跑在A股前面，差0.57到4.82个点。那天不是A股在炒，是港股更热。君正的A/H比价从1.4162缩到1.4094，一天缩0.48%——A股相对港股还少涨了一点。
这个对照实验也有边界，不能当铁证用。君正H股只有3128.73万股，占总股本6.08%，10月28日成交2.04亿港元，收盘市值24.86亿港元；A股那天成交72.75亿元。小池子做出18.94%，比大池子容易得多。所以港股确认的是方向和存在性——这份文件在两个市场都被读成了利好；它不确认幅度。
## 兆易是同行里最干净的一把尺子
兆易那天没有任何经营公告，上一条是10月28日的回购进展，三季报预约披露日排在10月30日。它当天的涨幅里没有财报。用它量，板块那一层在A股是7.77个点、在港股是9.91个点；君正减掉尺子，财报层在A股是10.60、在港股是9.03。两个市场给出同一个数。
## 麻烦在于，这10.6个点没法用财报质量解释
东芯10月27日也挂三季报：单季收入17.95亿元、环比增76.4%（君正31.1%），单季毛利率79.31%（君正65.93%），毛利率环比扩4.8个点（君正扩8.8个点），净利环比59.3%对61.5%。六个指标互有胜负，没有一项能撑起10个点的差距。
它当天的走法更说明问题。跳空高开5.71%，收跌6.94%，从最高62.29元杀到最低54.52元，一天之内高低差12.5%，收在全天最低附近。君正反过来：跳空11.25%，收涨18.37%，收在全天振幅83.8%的位置，缺口一分没回补。同样是隔夜读完文件的跳空，一个被砸掉，一个被追着买。
差的是位置。10月28日之前的20个交易日，君正跌35.25%，兆易跌33.11%，东芯跌33.74%，佰维跌30.32%，江波龙跌34.51%，普冉跌38.80%。整条链在10月被砍掉三成，基本面却在同一个10月往上走。弹药是共用的，扳机是独有的：10月27日东芯扣了一次，那天创业板指跌3.41%，市场不肯接；10月28日君正扣，指数涨3.40%，市场接住了。
## 扳机扣响之后，幅度由报表结构决定
单季收入比去年同期多19.58亿元，毛利额多16.97亿元——收入增量里86.7%直接落到毛利上。销售、管理、研发三项合计从3.46亿元到4.81亿元，只增39.0%，占收入的比例从29.15%压到15.30%。
这笔钱最终落在"营业成本"这一行的相对减速上，不是落在税金，也不是落在减值。刘强在同一个场合说得很直白：纯代工厂价格涨幅相对有限，总体代工成本上升已通过产品定价调整转嫁给客户。收入按新价结算，成本按旧价小幅上调，差额全部留在毛利里。
两头一夹，单季净利率从4.43%抬到45.11%。40.68个点的抬升里，毛利率贡献34.30，费用率摊薄贡献13.86，资产减值和所得税反向吃掉7.48。26倍的利润增长不是省出来的，是涨价弹性乘费用杠杆。这条杠杆的斜率完全由收入决定，所以被定价的不是这14.19亿元，是收入还能不能再涨一季。
## 毛利率最高的那条线，公司自己说要回落
| 产品线 | 1-9月收入(亿元) | 同比(%) | 毛利率(%) |
|---|---|---|---|
| 存储芯片 | 49.16 | 130.73 | 56.53 |
| 计算芯片 | 17.52 | 92.89 | 64.04 |
| 模拟芯片 | 4.18 | 11.70 | 49.28 |
三条线里最反直觉的是计算芯片：涨价喊的是存储，毛利率最高的却是它，64.04%，比存储的56.53%高7.5个点。原因写在同一份季报里——计算芯片受市场需求增长和存储原材料紧缺两方面影响。同一条涨价链，从存储颗粒流到计算芯片的成本，再从售价流出去，两头都记在这家公司账上。
董事长刘强在10月27日那场电话会上给这条线留了话：因缺货涨价带来的业绩提升是暂时的。也就是说，65.93%这个整体毛利率里，有一块来自公司自己预期要回落的产品线。
资产负债表还有一处对不上：合同负债0.616亿元、较上年期末降43.82%，原因是预收货款已结转——客户的钱在变成收入，不是订单在堆。同期资产减值损失0.949亿元、同比增68.09%，计提的是存货跌价和开发支出减值。涨价周期里还要计提存货跌价，说明不是所有料都在赚钱。
## 涨完18%还是8到10倍
总股本5.1495亿股，收111.98元，总市值576.6亿元。用三季度单季14.19亿元年化56.7亿元，10.2倍；用A股流通市值473.5亿元算，8.3倍。换成上半年净利11.98亿元年化，24.1倍；换成2025年全年3.76亿元，153倍。
涨了18.37%之后还是8到10倍。一个真被当成成长股重新定价的公司，不会停在这个数上。所以这18个点涨的不是新常态，是"这次比我以为的更极端"——市场一边抬价，一边已经把三季度的利润判成一次性。
最强的反驳在这里：8倍本来就是周期股顶部附近的正常倍数，拿它证明"市场不信可持续"，恰恰说明市场是对的。这个反驳成立。所以这里不用倍数说贵或便宜，只用它说一件事：估值体系没变，变的只是对周期高度的估计。反过来说，四季度毛利率若回落，利润只要掉四成五，8倍就自己变成15倍，那个过程不需要股价再跌。
## 什么情况算看错
- 10月30日兆易三季报（预约披露日2026-10-30，截至10月29日凌晨未披露）。如果它单季毛利率低于二季度的66.57%，7.77和9.91这把同行尺子就悬空，板块那一层要还回去。
- 如果四季度单季毛利率低于65.93%，"继续增长"这句话就废了。公司自己留了余地：部分产品涨幅已收窄，明年的价格无法给出明确预测。
- 如果A股与H股比价超过1.5（现在是1.41），说明港股先跑，A股那1.57个点的溢价要还。
来源：北京君正集成电路股份有限公司《2026年第三季度报告》，2026年10月28日，公告编号1225631045；《投资者关系活动记录表》，编号20261027，2026年10月28日，公告编号1225637336；《关于稳定价格行动、稳定价格期结束及超额配股权失效的公告》，2026年9月21日，公告编号1225573887；东芯半导体《2026年第三季度报告》，2026年10月27日，公告编号1225626412；兆易创新《关于以集中竞价交易方式回购A股股份的进展公告》，2026年10月28日，公告编号1225630578；港股与A股涨跌幅、成交额、A/H比价取交易所日线（2026/10/28），恒生指数、创业板指取指数日线；单季数据由定期报告累计数相减算得；市盈率为作者按总市值与单季净利年化计算。