安培龙为什么今天涨停，有什么概念和催化吗
- 累计口径：前三季营收9.48亿、增10.01%，归母981万、减86.58%，扣非901万、减85.42%，经营现金流-4193万（去年同期是+7315万）
- 今天收盘总市值100.06亿，滚动12个月归母净利2743万（2025年全年9075万，减2025年前三季7313万，加2026年前三季981万），对应364.79倍
- 宇树科技涨12.49%、市值2911亿，汉威科技涨11.79%，柯力传感涨6.62%，雷赛智能涨7.62%
- 而报表上能看见的方向都还朝下：毛利率同比-7.57个百分点、经营现金流-4193万、存货4.05亿比年初增16.6%、存货跌价多提891万
- 另一件是订单能见度：三季度末合同负债180.9万，比年初91.7万增97.3%，百分比好看，绝对额只有181万、占前三季营收0.19%——报表上没有机器人批量订单留下的预收款痕迹
安培龙为什么今天涨停，有什么概念和催化吗
Why did Amperex Technology Limited hit the daily limit today? What concepts and catalysts are there?
The desk answers已核实Walnut, from public filings
今天这个20cm不是三季报给的。三季报昨晚16:31才挂上巨潮，今早开盘60.07元，比前收60.69元还低1.02%，盘中最低59.48元、跌2.0%；9点30分到10点05分整个机器人板块被拉起来，它才从-2%冲到+15%，10点17分15秒封上涨停。业绩在这一天里没出过一分力。
先看当天涨的是谁。收盘全市场79只涨停、0只跌停，涨停归因表78条记录里54条带机器人、减速器、丝杠、传感这类标签。同样20cm封板的有豪恩汽电、伟创电气、恒工精密；宇树科技涨12.49%、市值2911亿，汉威科技涨11.79%，柯力传感涨6.62%，雷赛智能涨7.62%。反过来，真正做汽车传感器的保隆科技只涨3.14%，苏奥传感涨2.71%。这条对照最说明问题：今天资金买的不是"传感器"，是"人形机器人"四个字。安培龙是创业板，20cm的板正好给了情绪一个更大的容器。
那它身上的"机器人"到底是什么。9月30日那份投资者关系活动记录表，第7个问题问的就是机器人业务，公司原话：用于机器人的产品主要为力传感器，涵盖六维力传感器、力矩传感器、拉压力传感器，有"MEMS硅基半导体应变计结合玻璃微熔工艺"和"金属箔应变计"两条技术路线，"已累计向多家人形机器人、协作机器人及工业机器人相关上下游企业送样"，"部分客户已取得批量订单并实现了批量生产交付"。同一段里公司自己补了一句风险提示：机器人相关业务销售收入占公司整体营业收入比例较低，对公司报告期经营业绩不构成重大影响。
收入结构更直白。2026年上半年压力传感器3.01亿、占50.71%，同比只增2.91%；热敏电阻及温度传感器2.77亿、占46.76%，增10.38%；氧传感器及其他0.15亿、占2.53%，增49.12%。MEMS压力传感器5389.81万、增120.45%，是跑得最快的一块，但基数小。也就是说，它现在是一家汽车和家电传感器公司，"人形机器人"到收入之间隔着的不是一条线，是一个阶段：从送样到小批量。同花顺把485只股票装进"人形机器人"概念，安培龙是其中之一——这种宽度的标签，本来就不能用来判断谁真受益。
催化是一串，按日期摆出来。10月19日，工信部在国新办发布会上讲，今年上半年人形机器人整机产品突破400款、产量超4万台，预计全年突破10万台（这条我没拿到发布会文字实录，是21世纪经济报道、上海证券报、界面新闻同一天的一致转述）。10月21日，特斯拉三季度更新材料原文写着：Optimus production began in Fremont on our high-volume lines. The bots coming off this line are the first production-quality bots that we have manufactured. 海外整车厂从"研发验证"走到"产线下线"，这是这轮最硬的一条。10月22日，宇树科技王兴兴在世界物理AI大会上讲"物理AI已经看到被击穿的苗头，很可能最近1—2年或者2—3年"。近日首批六项人形机器人国家标准获批发布、2027年4月1日起实施，商务部等九部门印发《关于加快推进科技兴商的意见》提具身智能在商贸领域落地。10月23日商务部数据，前三季度重点平台具身智能机器人销售额增长22.6%。10月27日马斯克谈机器人内存需求，那条点的是存储，不是本体。落到个股，只有10月27日盘后那份三季报和一份计提减值公告。而公司自己在10月22日的异动公告里写得很清楚：主营业务未发生变化，不存在应披露而未披露的重大事项。
三季报的真实样子要拆单季看，累计数会把变化盖住。单季营收3.55亿、同比增15.18%，至少是2024年以来单季最高；单季归母631万、同比减79.64%；单季毛利率22.07%，去年同期29.64%。累计口径：前三季营收9.48亿、增10.01%，归母981万、减86.58%，扣非901万、减85.42%，经营现金流-4193万（去年同期是+7315万）。
利润掉的账能一笔一笔对上。前三季毛利2.15亿，去年同期2.45亿，少3040万；研发费用8562万，多花2096万、增32.42%；财务费用1647万，多786万，公司说是汇兑损失；资产减值多提467万，其中存货跌价891万；平湖厂房搬迁报废让营业外支出多311万。这五项加起来6700万，占利润总额同比减少7842万的85%。剩下的缺口靠所得税抵——所得税从去年同期的+231万变成-1252万，一来一去补了1484万。这块钱是税上省出来的，不是经营赚出来的。
环比是另一回事，也是多头唯一能抓的东西。单季营收Q1 2.62亿、Q2 3.32亿、Q3 3.55亿；单季毛利率25.64%、20.94%、22.07%；单季归母309万、41万、631万。Q3确实比Q2好，收入创新高、毛利率环比回来1.13个百分点。但这是从一个很低的坑底往回爬，同比还差着79.64%。
同行对照不能一句"行业都这样"带过。已经出三季报的可比公司，单季归母：保隆科技-0.12亿（去年同期+0.63亿，当季营收还增30%），苏奥传感0.05亿、减83.2%，东华测试0.15亿、减38.1%，四方光电0.32亿、增38.1%。金银铜高价、车企年降是整个汽车传感器环节的共同压力，公司记录表里自己就写了"大宗原材料（金、银、铜等）市场价格持续高位运行""下游客户年度降价诉求"。但四家能分化成这样，说明压力之外还有各自的产品结构和客户结构。安培龙的单季毛利率同比掉7.57个百分点，比苏奥的2.18个点、东华的3.71个点、保隆的1.81个点都大——这一档它不是被行业拖下去的，是自己站得最靠下。
涨了多少、定价到哪一步。10月19日收盘46.88元到今天72.83元，6个交易日涨55.3%。10月22日龙虎榜上三个机构专用席位合计净买入1.33亿元，深股通席位净买入约0.26亿元，同一天两家招商证券深圳营业部合计净卖出约0.50亿元——机构在买、本地营业部在卖，这和"纯情绪票"的印象并不完全一样。今天换手15.49%、成交8.07亿、封单1.75亿、盘中开板1次，筹码换得很快。
估值这块把口径说清。今天收盘总市值100.06亿，滚动12个月归母净利2743万（2025年全年9075万，减2025年前三季7313万，加2026年前三季981万），对应364.79倍。公司在10月22日异动公告里自己提示过：滚动市盈率140.62倍，所处行业37.69倍。十天时间，分母从五千万级掉到两千七百万级，同样六十块出头的股价，PE从140倍变成365倍。这个位置买的不是今年的利润，也不是明年的利润，是一个还没出现在报表里的量产年份。
还有两件事没被定价。一件是12月19日，首发前限售股解禁5154.76万股、占总股本37.52%（邬若军4051.02万股、瑞航投资721.66万股、黎莉382.08万股），五十多天后的硬约束，现在的价格里看不到折价。另一件是订单能见度：三季度末合同负债180.9万，比年初91.7万增97.3%，百分比好看，绝对额只有181万、占前三季营收0.19%——报表上没有机器人批量订单留下的预收款痕迹。
反方最强的那套说法我也摆出来。他会说你盯着同比看的是后视镜：单季收入历史新高、毛利率环比回升、归母从41万回到631万、研发多花32%是主动投入不是失控，MEMS压力增120%、玻璃微熔4月量产、泰国工厂进了客户验证、机器人从送样走到小批量——这正是投入期见底的形态，2027年费用摊薄叠加机器人放量，利润弹性会很大，用当期364倍PE去衡量它本身就用错了尺子。
这套说法有道理，但它缺一个数字：机器人到底贡献了多少收入。三季报没有单独披露，公司只说"占比低"。所以"放量"目前是预期，不是报表事实。而报表上能看见的方向都还朝下：毛利率同比-7.57个百分点、经营现金流-4193万、存货4.05亿比年初增16.6%、存货跌价多提891万。预期可以很贵，但预期不能当成已经发生的事来计价。
什么情况算我看错。Q4单季毛利率回到25%以上，同时2026年年报把机器人相关收入单独列出来、占比过5%，那"这只是主题"这个判断就错了。12月19日解禁前后大股东一份没减、股价也没被压下去，"解禁是硬约束"这条要撤回。机器人板块退潮——人形机器人指数连跌三天、每天涨停家数掉到20家以下——它还能独立走强，那"今天涨的是板块"这个归因就错了。
来源：
深圳安培龙科技股份有限公司《2026年第三季度报告》，2026年10月28日，公告编号2026-077
深圳安培龙科技股份有限公司《关于2026年前三季度计提资产减值准备的公告》，2026年10月28日，公告编号2026-078
深圳安培龙科技股份有限公司《关于股票交易异常波动的公告》，2026年10月22日，公告编号2026-076
深圳安培龙科技股份有限公司《投资者关系活动记录表》，2026年9月30日，编号2026-010
Tesla, Inc.《Q3 2026 Update》（Form 8-K，Items 2.02、9.01），2026年10月21日，accession 0001628280-26-066738
保隆科技《2026年第三季度报告》，2026年10月28日，巨潮公告ID 1225631080
苏奥传感2026年第三季度报告，2026年10月17日；四方光电《2026年第三季度报告》，2026年10月24日；东华测试2026年第三季度报告，2026年10月22日
深圳证券交易所行情、涨停池与龙虎榜数据，2026年10月28日收盘（龙虎榜为2026年10月22日）
工业和信息化部2026年10月19日国新办新闻发布会数据，经21世纪经济报道、上海证券报、界面新闻2026年10月28日报道转述
Today's 20cm limit was not given by the third-quarter report. The third-quarter report was only posted on Cninfo at 16:31 last night. This morning it opened at 60.07 yuan, still 1.02% lower than the previous close of 60.69 yuan; the intraday low was 59.48 yuan, down 2.0%; from 9:30 to 10:05 the entire robotics sector was pulled up, and only then did it surge from -2% to +15%, sealing the limit up at 10:17:15. Earnings did not contribute even a fraction of effort on this day.
First look at who rose that day. At the close, the whole market had 79 limit-ups and 0 limit-downs; among the 78 records in the limit-up attribution table, 54 carried labels such as robotics, reducers, lead screws, and sensing. Those also sealing the 20cm limit included Haun Auto Electronics, Weichuang Electric, and Henggong Precision; Unitree Technology rose 12.49% with a market value of 291.1 billion yuan, Hanwei Technology rose 11.79%, Keli Sensing rose 6.62%, and Leadshine Intelligent rose 7.62%. Conversely, Baolong Technology, which actually makes automotive sensors, rose only 3.14%, and Suao Sensing rose 2.71%. This contrast is the most telling: what capital bought today was not "sensors" but the four words "humanoid robot." Anpeilong is on ChiNext, and the 20cm board precisely gave sentiment a larger container.
So what exactly is the "robot" on it. In the investor relations activity record sheet dated September 30, the seventh question asked precisely about the robotics business. The company's original words: the products used for robots are mainly force sensors, covering six-axis force sensors, torque sensors, and tension/compression force sensors, with two technology routes—"MEMS silicon-based semiconductor strain gauges combined with glass micro-fusion process" and "metal foil strain gauges"—and it "has cumulatively sent samples to multiple upstream and downstream enterprises related to humanoid robots, collaborative robots, and industrial robots," and "some customers have obtained batch orders and achieved mass production delivery." In the same paragraph the company itself added a risk warning: revenue from robot-related business accounts for a relatively low proportion of the company's overall operating revenue and does not constitute a material impact on the company's operating performance for the reporting period.
The revenue structure is even more straightforward. In the first half of 2026, pressure sensors were 301 million yuan, accounting for 50.71%, up only 2.91% year-on-year; thermistors and temperature sensors were 277 million yuan, accounting for 46.76%, up 10.38%; oxygen sensors and others were 15 million yuan, accounting for 2.53%, up 49.12%. MEMS pressure sensors were 53.8981 million yuan, up 120.45%, the fastest-growing piece, but on a small base. That is to say, it is now an automotive and home appliance sensor company, and what lies between "humanoid robot" and revenue is not a line but a stage: from sample delivery to small-batch. Tonghuashun packs 485 stocks into the "humanoid robot" concept, and Anpeilong is one of them—a label of that breadth simply cannot be used to judge who truly benefits.
The catalysts are a string of them, laid out by date. On October 19, the Ministry of Industry and Information Technology said at the State Council Information Office press conference that in the first half of this year humanoid robot complete-machine products surpassed 400 models and output exceeded 40,000 units, with the full year expected to surpass 100,000 units (I did not obtain the transcript of the press conference; this is the consistent paraphrase on the same day by 21st Century Business Herald, Shanghai Securities News, and Jiemian News). On October 21, Tesla's third-quarter update materials stated verbatim: Optimus production began in Fremont on our high-volume lines. The bots coming off this line are the first production-quality bots that we have manufactured. An overseas automaker moving from "R&D validation" to "rolling off the production line" is the hardest item in this round. On October 22, Unitree Technology's Wang Xingxing said at the World Physical AI Conference that "physical AI has already shown signs of being broken through, very likely in the next 1–2 years or 2–3 years." Recently the first batch of six humanoid robot national standards was approved and released, effective April 1, 2027, and the Ministry of Commerce and eight other departments issued the "Opinions on Accelerating the Promotion of Science and Technology in Commerce," mentioning the implementation of embodied intelligence in the commerce sector. On October 23, Ministry of Commerce data showed that in the first three quarters sales of embodied intelligent robots on key platforms grew 22.6%. On October 27, Musk talked about robots' memory demand; that item points to storage, not the body. Down to the individual stock, there were only the third-quarter report after the close on October 27 and an impairment provision announcement. And the company itself wrote very clearly in its October 22 abnormal movement announcement: the main business has not changed, and there are no undisclosed material matters that should be disclosed.
The true shape of the third-quarter report must be broken down by single quarter; cumulative figures cover up the changes. Single-quarter revenue was 355 million yuan, up 15.18% year-on-year, at least the highest single quarter since 2024; single-quarter net profit attributable to parent was 6.31 million yuan, down 79.64% year-on-year; single-quarter gross margin was 22.07%, versus 29.64% in the same period last year. Cumulative basis: first three quarters revenue was 948 million yuan, up 10.01%, net profit attributable to parent was 9.81 million yuan, down 86.58%, excluding non-recurring items it was 9.01 million yuan, down 85.42%, and operating cash flow was -41.93 million yuan (versus +73.15 million yuan in the same period last year).
The profit decline can be reconciled item by item. Gross profit for the first three quarters was 215 million yuan, versus 245 million yuan in the same period last year, a decrease of 30.4 million yuan; R&D expenses were 85.62 million yuan, up 20.96 million yuan, or 32.42%; financial expenses were 16.47 million yuan, up 7.86 million yuan, which the company attributes to exchange losses; asset impairment was 4.67 million yuan higher, including 8.91 million yuan of inventory write-downs; the scrapping from the Pinghu plant relocation added 3.11 million yuan to non-operating expenses. These five items add up to 67 million yuan, accounting for 85% of the 78.42 million yuan year-on-year decrease in total profit. The remaining gap is covered by income tax—income tax went from +2.31 million yuan in the same period last year to -12.52 million yuan, a swing that made up 14.84 million yuan. This money was saved on taxes, not earned from operations.
The quarter-on-quarter picture is a different matter, and it is the only thing the bulls can latch onto. Quarterly revenue was 262 million yuan in Q1, 332 million yuan in Q2, and 355 million yuan in Q3; quarterly gross margin was 25.64%, 20.94%, and 22.07%; quarterly net profit attributable to parent was 3.09 million yuan, 0.41 million yuan, and 6.31 million yuan. Q3 is indeed better than Q2, with revenue hitting a record high and gross margin recovering 1.13 percentage points quarter-on-quarter. But this is climbing back from a very low trough, and it is still down 79.64% year-on-year.
Peer comparison cannot be waved away with a single line about "the whole industry being like this." Among comparable companies that have already reported third-quarter results, quarterly net profit attributable to parent: Baolong Technology -12 million yuan (versus +63 million yuan in the same period last year, with quarterly revenue still up 30%), Suao Sensor 5 million yuan, down 83.2%, Donghua Testing 15 million yuan, down 38.1%, and Siwei Optoelectronics 32 million yuan, up 38.1%. High gold, silver, and copper prices and annual price cuts by automakers are common pressure across the entire automotive sensor segment, and the company itself wrote in its records that "market prices of bulk raw materials (gold, silver, copper, etc.) continue to run at high levels" and "downstream customers' annual price-cut demands." But the fact that these four companies diverged so much shows that beyond the pressure, each has its own product mix and customer mix. Amperon's quarterly gross margin fell 7.57 percentage points year-on-year, a larger drop than Suao's 2.18 points, Donghua's 3.71 points, and Baolong's 1.81 points—at this level it was not dragged down by the industry; it is standing the lowest.
How much it has risen, and where pricing has gotten to. From a close of 46.88 yuan on October 19 to 72.83 yuan today, it rose 55.3% in 6 trading days. On the October 22 dragon-tiger list, three institutional special seats had combined net buying of 133 million yuan, the Shenzhen Stock Connect seat had net buying of about 26 million yuan, and on the same day two China Merchants Securities Shenzhen branches had combined net selling of about 50 million yuan—institutions buying while local branches sell, which is not entirely consistent with the impression of a "pure sentiment stock." Today turnover was 15.49%, trading value 807 million yuan, sealing orders 175 million yuan, and the board opened once intraday; chips are changing hands very quickly.
On valuation, let me make the basis clear. Today's closing total market capitalization is 10.006 billion yuan, and trailing 12-month net profit attributable to parent is 27.43 million yuan (full-year 2025 of 90.75 million yuan, minus the first three quarters of 2025 of 73.13 million yuan, plus the first three quarters of 2026 of 9.81 million yuan), corresponding to 364.79 times. In its October 22 abnormal-movement announcement, the company itself noted: trailing price-to-earnings ratio of 140.62 times, versus 37.69 times for its industry. In ten days, the denominator fell from the fifty-million level to the twenty-seven-million level, and at the same share price of just over sixty yuan, PE went from 140 times to 365 times. What is being bought at this level is not this year's profit, nor next year's profit, but a mass-production year that has not yet appeared in the financial statements.
There are two more things that have not been priced in. One is December 19, when 51,547,600 pre-IPO restricted shares will be unlocked, accounting for 37.52% of total share capital (Wu Ruojun 40,510,200 shares, Ruihang Investment 7,216,600 shares, Li Li 3,820,800 shares), a hard constraint more than fifty days away, and no discount is visible in the current price. The other is order visibility: contract liabilities at the end of the third quarter were 1.809 million yuan, up 97.3% from 0.917 million yuan at the beginning of the year; the percentage looks good, but the absolute amount is only 1.81 million yuan, 0.19% of first-three-quarter revenue—there is no trace in the statements of advance receipts left by batch robot orders.
Let me also lay out the strongest bear case. It would say you are looking at year-on-year figures in the rearview mirror: quarterly revenue at a record high, gross margin recovering quarter-on-quarter, net profit attributable to parent back from 0.41 million yuan to 6.31 million yuan, R&D spending up 32% as proactive investment rather than loss of control, MEMS pressure up 120%, glass micro-fusion mass production in April, the Thailand plant entering customer validation, and robots moving from sampling to small batches—this is exactly the shape of an investment period bottoming out, and in 2027, with expenses diluted and robot volume ramping, profit elasticity will be large; measuring it with the current 364 times PE is itself using the wrong ruler.
This argument makes sense, but it lacks one number: how much revenue robots actually contribute. The third-quarter report does not disclose it separately; the company only says "the proportion is low." So "volume ramp" is currently an expectation, not a statement fact. And every direction visible in the statements is still downward: gross margin -7.57 percentage points year-on-year, operating cash flow -41.93 million yuan, inventory of 405 million yuan up 16.6% from the beginning of the year, and inventory write-downs 8.91 million yuan higher. Expectations can be very expensive, but expectations cannot be priced as if they have already happened.
Under what circumstances would I be considered to have judged wrongly? If the Q4 single-quarter gross margin returns above 25%, and at the same time the 2026 annual report lists robot-related revenue separately with its share exceeding 5%, then the judgment that "this is only a theme" would be wrong. Around the December 19 lock-up expiry, if the major shareholder did not reduce a single share and the stock price was not suppressed, then the point that "lock-up expiry is a hard constraint" must be retracted. If the robot sector retreats — the humanoid robot index falls for three consecutive days and the number of limit-up stocks drops below 20 per day — and it can still strengthen independently, then the attribution that "what rose today is the sector" would be wrong.
Sources:
Shenzhen Ampron Technology Co., Ltd. "2026 Third Quarter Report," October 28, 2026, announcement no. 2026-077
Shenzhen Ampron Technology Co., Ltd. "Announcement on the Provision for Asset Impairment for the First Three Quarters of 2026," October 28, 2026, announcement no. 2026-078
Shenzhen Ampron Technology Co., Ltd. "Announcement on Abnormal Fluctuations in Stock Trading," October 22, 2026, announcement no. 2026-076
Shenzhen Ampron Technology Co., Ltd. "Investor Relations Activity Record Form," September 30, 2026, no. 2026-010
Tesla, Inc. "Q3 2026 Update" (Form 8-K, Items 2.02, 9.01), October 21, 2026, accession 0001628280-26-066738
Baolong Technology "2026 Third Quarter Report," October 28, 2026, CNINFO announcement ID 1225631080
Suo'ao Sensing 2026 Third Quarter Report, October 17, 2026; Sifang Optoelectronics "2026 Third Quarter Report," October 24, 2026; Donghua Testing 2026 Third Quarter Report, October 22, 2026
Shenzhen Stock Exchange market, limit-up pool and Dragon-Tiger List data, October 28, 2026 close (Dragon-Tiger List dated October 22, 2026)
Ministry of Industry and Information Technology data from the State Council Information Office press conference on October 19, 2026, as relayed in reports by 21st Century Business Herald, Shanghai Securities News, and Jiemian News on October 28, 2026
三波点火性质不一样。10月20日46.88到52.31元、+11.58%，对得上10月19日工信部发布会；10月22日53.34到61.13元、+14.60%，当晚就挂出异动公告，对得上10月21日特斯拉三季度材料里那句Optimus已在Fremont高量产线下线；10月23日再+10.91%到67.80元，随后两天回调到60.69元。今天是第三波。前两波各有硬事件托底，今天没有：10月27日盘后三季报单季归母-79.64%，10月22日公司自己写"主营业务未发生变化"。今天成交8.07亿元、换手15.49%，低于10月22日的20.10%和10月23日的21.92%——缩量新高，推价的不是新钱，是没人卖。
催化按能不能落到报表分三档，混着看会出错。落到报表的：零，三季报里没有一行叫机器人。落到资本开支的只有一项：定增募投"力传感器产线建设项目"，投资总额6250万元、拟投募集资金6040万元；半年报原文"在深圳总部建成了基于玻璃微熔技术与MEMS技术融合的力传感器生产线"，上海另有一条金属应变计线。这句能对上账：三季度末在建工程9905.7万元、年初9409.4万元，九个月只增加496万元，同期固定资产从8.449亿增到9.066亿、购建长期资产付现1.139亿（+31.36%）——线已建成转固，产能就位。只能落到情绪的：六项国标2027年4月1日实施，对2026年年报贡献0个月，它定的是标准，不是订单。
卡点不在传感器厂这一层。力传感器的生意是批量小、标定和一致性是门槛，客户导入周期长；复用得上的是工艺，玻璃微熔本就是做压力传感器的那套设备；复用不上的是客户，汽车和机器人是两套认证。真正卡在下一层：整机厂没量产，传感器厂就拿不到批量订单，节奏由整机厂决定，安培龙再多建一条线也不能把收入提前做出来。这就是为什么"力传感器生产线已建成"和"相关业务销售收入占整体营业收入比例较低"能同时成立，也是为什么9月30日记录表那句"部分客户已取得批量订单并实现了批量生产交付"不能翻译成收入——"批量"修饰的是订单，不是亿元。
上限有硬边界，不用猜。半年报分产品只有两行：热敏电阻及温度传感器2.775亿元、压力传感器3.009亿元，合计5.784亿元；营业收入5.934亿元，差额1501.19万元就是氧传感器、力传感器和其他的全部，占2.53%。力传感器这半年收入上限就是1501万元，还得扣氧传感器。100.06亿元对应1501万元是666倍；半年报引的QY Research说2025年全球六维力传感器市场4.09亿美元、约29亿元，这个上限只相当于其0.5%。同一张表里压力传感器毛利率25.87%、热敏20.00%，力传感器连毛利率都没单列，没法与这两条对比。
乘数拆开，分歧点不是大家以为的那个。364.79倍是100.06亿元除以滚动12个月归母2743.0万元；压到60倍需要净利1.6677亿元、多出1.3934亿元。按公司2025年7.66%的净利率要新增收入18.2亿元，就算给15%的净利率也要9.3亿元——前三季总收入才9.48亿元，等于再造两个自己。换成台数：半年报引的GGII口径，2030年全球人形机器人销量接近34万台；单台力传感器价值量若5000元，全球盘子17亿元，18.2亿要拿下107%；若2万元，盘子68亿元，要占26.8%。这个价值量是我的假设，公司没披露，六维力和关节力矩口径也不统一——决定364倍是贵是便宜的，正是这个没人能定的乘数。
反方那张牌我认，比多数人想的硬：8月13日到账的定增是竞价发行，底价48.12元，最后发在57.49元、是底价的119.47%，13家现金认购5.444亿，其中诺德1.239亿、鹏华0.8亿、财通0.789亿、摩根士丹利0.34亿。加上单季收入3.55亿元创新高、毛利率环比回来1.13个百分点、归母从41万回到631万元，"投入期见底"确实有报表支撑。我的回应只有一句：那批钱成本57.49元，现价比它高26.7%，锁到2027年2月；机构买的是"线已建成、行业没人量产"这张期权，能定很高的价，但不是订单。要证明他们对，三样出现一样就够：力传感器单列、合同负债跳升、单季毛利率回到25%以上。现在都没有——合同负债180.9万元，占前三季营收0.19%。
没定价的有三件。一是供给日历：流通股7637.50万股只占总股本1.3739亿股的55.6%；今年12月18日（上市满36个月）解禁5154.76万股，流通股变1.2792亿股、扩大67.5%，流通市值55.6亿→93.2亿元；2027年2月定增946.95万股再解禁到100%。日成交8亿的票三个月可交易盘子扩大67%，这个折价在K线上看不到。二是质押：邬若军持股4051.02万股、已质押1071.5万股（占其持股26.45%），其中60万股是10月9日补押给云南国际信托，公告写"补充质押"；质押物100%是限售股，12月18日前卖不掉，之后既是筹码也是抵押品——7月质押、7月两次解押、10月再补押，一年用了四回。三是"具身智能"的股权布局：江苏鼎汇具身智能机器人创新中心300万元、无锡瑞知微600万元、海纳微112万元已按0元计价（清算中）；9月那个4250万元的创投基金是LP、不并表——都是姿态，不是资产。
位置给坐标。今天72.83元，在8月17日高点75.00元下方3%，7月3日盘中到过86.66元；四个月里两次40%级回撤——86.66元到10月15日最低43.90元跌49%，75.00元到同一个43.90元跌41%。两次都从这个区域掉头，掉头时"行业无大规模量产"一个字没变；这次不同在12月18日解禁等着。
什么情况下算我看错，四条。一是2026年年报把力传感器单列且收入过亿、占营收8%以上，"这只是主题"就作废。二是解禁后一个月内邬若军、瑞航投资、黎莉三家一份没减、质押也没再加，"供给是硬约束"撤回。三是连续两个季度单季毛利率回到25%以上、合同负债从180.9万元升到5000万元以上，"订单未到"这条就错了。四是人形机器人指数连跌三天、涨停家数掉到20家以下它还能独立走强，"今天涨的是板块"这个归因就错了。反向留一条：2027年2月定增解禁前，若有第三方口径把单台力传感器价值量证实到2万元以上，"分母不够"要放宽。
来源：
安培龙《2026年第三季度报告》，2026-10-28，2026-077
安培龙《2026年半年度报告》，2026-08-28（分产品收入表、力传感器产线、QY/GGII数据、其他权益工具投资）
安培龙《向特定对象发行股票发行情况报告书》，2026-08-17（底价48.12元、发行价57.49元、13名对象、六个月限售期）
安培龙《关于调整募集资金投资项目金额的公告》，2026-09-03，2026-063
安培龙《关于控股股东、实际控制人部分股份补充质押的公告》，2026-10-12，2026-073
安培龙关于控股股东部分股份质押及解除质押的公告，2026-07-07、07-14、07-24
安培龙《关于与专业投资机构共同投资设立创业投资基金的公告》，2026-09-03，2026-061
安培龙《关于股票交易异常波动的公告》，2026-10-22，2026-076
安培龙《投资者关系活动记录表》，2026-09-30，2026-010
Tesla, Inc.《Q3 2026 Update》（Form 8-K，Items 2.02、9.01），2026-10-21，accession 0001628280-26-066738
深圳证券交易所行情与成交数据，2026-07-01至2026-10-28
工业和信息化部2026-10-19国新办发布会数据，经21世纪经济报道、上海证券报、界面新闻同日报道转述