文 | 陆玖商业评论
复购收入含金量上升，曾经仅凭样机、概念故事就能拿到高估值的时代已经过去。把内卷当作主要战场，很难跑出真正的商业闭环，头部企业需要寻找新的破局路径，这步棋，甚至可以比新能源汽车走的更早。
智元也是很拼了，不拼又不行。
10月21日，智元宣布10月22日起连续七个工作日发布七项成果，覆盖数据集、作业平台、重载机器人G2Max、具身Agent框架GenieMaster、仿真平台、基座大模型与群体经验学习范式。
这有一个大背景，资本市场对具身智能的态度，在今年初和今年末，转了一个180°的大弯。前期整个具身智能赛道都挤破头冲向资本市场，时下大多的IPO则成了扑朔迷离的事件。
具体到智元身上，作为全球头部玩家，据报道，智元创新近期启动新一轮融资，市场传出估值约500亿元，但IPO已暂缓。
一个细节是，二级市场内人形机器人玩家的表现有些拉胯，使得资本开始从狂热变得耐心。大门已经不再对概念故事大开绿灯。此前。宇树登陆科创板后市值近乎腰斩、梅卡曼德港股上市超高认购却首日破发，接连上演的市场案例摆在眼前。
不过，暂缓IPO不等于放弃资本化。这更像是一次审时度势的选择：在行业估值泡沫出清的节点，与其仓促登台接受市场放大镜审视，不如退回赛道内部打磨业务基本面。原因是，具身智能的资本逻辑已经发生本质切换，市场评判标尺从样机、概念、故事，转向订单质量、客户复购、商业化真实成色。
多卖一台真机器，比多一分的股票价格，好像更值钱。
宇树在资本市场扑街，智元也变聪明了？
陆玖商业评论注意到，多位业内人士向外界透露，此前市场已多次传出智元赴港上市节奏放缓的消息，若《财经》报道的新一轮融资属实，智元在今年内无法完成IPO。
这不只是智元的问题，它是整个产业发展周期的一个缩影。过去，“狂飙”是近两年具身智能赛道一个典型的代表词，人形机器人行业亦弥漫着一种普遍共识：只要贴上人形机器人标签，就能复制芯片、半导体行业批量上市的资本盛宴。
仅仅只是因为上了春晚，具身智能火得有些突然，一级市场愿意为远期产业想象支付高额溢价，样机发布会、炫酷的演示视频，就足以撑起企业的估值叙事，这是早期的泡沫。
人形机器人产业发展至今确有在降温，有的被两场二级市场行情狠狠打破。
A股人形机器人第一股宇树科技8月登陆科创板，盘中市值一度突破4400亿元，短短数个交易日市值近乎腰斩，超2000亿市值蒸发。港股市场，机器人零部件厂商梅卡曼德公开发售获得约3835倍超额认购，刷新港股机器人赛道认购纪录，上市首日直接破发，后续股价持续走弱。
接连发生的行情变化，在传递一个清晰信号：资本已经从狂热资本转向耐心资本。二级市场投资者恐怕不想单纯为人形机器人概念买单，企业手里真实订单、市场化复购能力，权重被放到前所未有的高度。
对于包括智元在内的企业，按照港交所特专科技18C章的规则，企业递交上市申请之后，如果挂牌前真的完成新一轮融资，在未拿到交易所豁免的前提下，融资交割完成后必须等待120天才可以推进挂牌流程。
新融资客观拉长了其他未IPO的企业的上市时间表。但抛开规则层面的客观限制，行业环境的改变，是不可忽视的底层诱因。
毕竟，监管审核门槛已经明显抬升。港交所对外释放信号，计划在明年上半年针对18C特专科技上市制度开展咨询，其中就包含市值门槛的检讨，这会直接影响排队当中数十家具身智能企业的命运。同样，A股科创板6月修订发行上市规定，将机器人、具身智能列为重点支持方向的同时，明确从严把控上市准入，严守财务真实性底线，规避带病申报的现象。
Wind统计数据显示，港交所有超过50家具身智能赛道企业正在排队候场。但陆玖商业通过梳理后发现，产业链上的市场已经形成现实分化：上游零部件厂商上市通路相对顺畅，珞石机器人、仙工智能、梅卡曼德、本末科技已经先后完成港股挂牌；人形整机本体企业闯关阻力明显更大，短期内恐怕很难见到整机企业敲钟的案例。
因为高开低走的局面已是前车之鉴。
理论上，招股书是一份完全公开的考卷，所有收入结构、订单来源、客户构成，全部摊开在投资者、行业同行的视野之下。但赛道内已经出现公开争议，市场开始区分一笔收入来自市场化商业客户，还是哪里？
在这种背景下，想上市的企业的招股书如果被爆出真实收入含金量不够高，又将受到外界非议。显然，很少有企业愿意在准备不够充分的时候，直面来自资本市场、投资者以及社会大众对收入真实性的公开拷问。暂缓IPO，留出窗口期打磨业务指标，成为后面一些玩家的现实选择。
看来，即使是大厂也不得不为需求奔波，渐渐地，有理想相信在今后，比起抢时间敲钟，业务指标能不能经得起审计、能不能经得住二级市场的审视，变得更加重要。
既然资本市场的窗口已经发生改变，但外部只是压力来源，真正影响一家企业能走多远的，依旧在自身业务底盘之中。
智元的对手，在友商还是在自身含金量
从人形机器人行业竞争格局来看，智元与宇树构成赛道双寡头。
SAG机构2026上半年统计数据，智元8400台、宇树5900台的出货量，两家合计拿下全球超七成市场份额，其余中小厂商单家上半年出货量均未能突破千台，头部和腰部企业的规模鸿沟已经拉开。
宇树此前实现过盈利。表面看，智元最大对手是同赛道的一众头部友商。
但穿透行业表象，全赛道共同面临的发展堵点，同时也是压在智元身上的核心考题。真正的考验，来自企业自身业务含金量。
整个行业都正在被审视收入的真实成色。人形机器人场景应用联盟统计，2026上半年国内公开人形机器人中标项目共218个，总规模17.23亿元。其中教育科研机构采购占比55.5%，政府及国资平台占20.6%，真正来自工业企业订单仅有21.1%。而工业企业订单内部，大部分还停留在样机测试、展厅展示、实训试点，并非工厂产线常态化商用采购。
假设大量订单带有科研项目属性，预算跟随项目周期波动，很难形成持续性复购。一旦项目结束，对应营收就存在波动风险，这也是当下监管问询、二级市场投资人重点关注的风险点。这不是某一家企业的特例，而是绝大多数人形整机厂商共同面对的行业现实。
相比之下，智元有利好的一面，例如，预计11月2日至3日，6台G2机器人将进入真实便利店24小时直播作业，完成全栈技术的真实场景验证。
另外在2026年智元合作伙伴大会上，智元创始人兼CEO（首席执行官）邓泰华披露了智元成立以来的营收状况：“从成立第一年的30万元，到第二年的6000多万元，再到去年（2025年）突破了10亿元的营收。每年实现20倍的增长。今年的目标会再进一步实现数倍的增长。2027年要跨越100亿元的营收。”
只不过，这些不等于产品已经走向完全成熟。统计口径之下的出货，包含轮式、半尺寸机型，全尺寸人形机器人规模化部署还有提升空间。
国内大量整机企业，减速器、伺服电机、控制器等核心元器件对外采购占比偏高，供应链高度趋同，样机层面很容易做出能行走、能表演动作的机器人，但产品差异化更多停留在外观与演示动作，真实非结构化工业场景的作业能力，还需要海量工况数据打磨迭代。所以，即使是智元，设备可以日渐批量部署，但从演示效果到更稳定干活，中间隔着不小距离。
公司CEO邓泰华曾表示“不急于追求净利润”，他还强调，智元现在的策略是要保持研发领域饱和投入，来构建产品竞争力。
重研发赛道之下，企业的现金流压力客观存在。硬件迭代、算法训练、团队人力、市场拓展，每一块都持续消耗资金。智元尚未实现盈利，对外部融资具备一定依赖。
回过头来看，同行之间的出货比拼、市场份额争夺，只是表层的竞争。订单结构优化程度、产品实际落地能力、收敛亏损的路径，这些业务底盘层面的命题，才是智元需要跨越的内部关卡。
相较于国内头部玩家愈演愈烈的厮杀，陆玖商业评论观察到了另外一种趋势，把内卷当作主要战场，很难跑出真正的商业闭环，头部企业需要寻找新的破局路径，这步棋，甚至可以比新能源汽车走的更早。
对内卷同行，不如提前对外出海
加快向外冲这个推测并非没有道理。
回顾新能源汽车产业发展路径，国内厂商大多经历先深耕本土市场，积累技术与规模，再向外出海的发展顺序。落到具身智能赛道，行业却出现了不一样的可能性：国产人形机器人拥有供应链、成本制造优势，出海可以更早成为重要增长引擎。
就大盘数据而言，数据显示，今年一季度，中国人形机器人出口同比增长210%，彭博社10月14日刊文称，中国有望引领价值5万亿美元的人形机器人市场。
具体到智元身上，智元机器人合伙人、联席总裁姜青松介绍，公司计划2026年海外销售占比达到30%以上，未来有望超过50%。
上述数据意味着，向内打磨业务基本面，向外抢夺海外增量市场，可以成为头部整机企业更现实的出路，出海有机会成为这套解法当中至关重要的第二曲线。
中国人形机器人出海有充足的优势，摩根士丹利研报给出判断，人形机器人产业发展阶段等同于2018至2020年新能源汽车，未来五年将迎来全球出货量百倍增长，中国厂商牢牢占据产业链主导地位。新能源车依靠能源替代逻辑打开海外市场，人形机器人依靠劳动力替代刚需撬动全球需求，市场空间、扩张速度、产业链话语权全面占优，出海浪潮爆发力更强。
劣势在于，外界市场是否会规模性地接受产品的准入，这考验着包括智元在内的产品合规能力，同时，来自竞争市场的第一堵高墙亦不易忽视。
据可靠消息，随着特斯拉透露其第三代人形机器人Optimus（擎天柱）已做好大规模量产准备，市场目光再度聚焦这一企业，如同前段时间特斯拉飞上天一样，国产人形机器人品牌不能忽视来自特斯拉在全球市场的竞争压力。马斯克甚至直言，Optimus人形机器人和Cybercab将构成特斯拉绝大部分价值。
第二堵墙是，产业化之路并非坦途，硬件可靠性、物理AI算法泛化能力、成本控制、行业合规标准及数据闭环建设，仍是智元等人形机器人产业发展的核心约束。
回过头来看，放眼港股排队队伍，技术演示和一门可持续的生意，中间隔着很长的产业化路程。与其继续在国内存量赛道内卷厮杀，不如把资源投向产品打磨、投向全球市场，去争取来自海外市场的真实商业客户，只是这条路机会颇大荆棘也不会少。
智元被爆暂缓IPO这件事，算不上行业拐点，一级市场依旧愿意为人形机器人的长期产业空间下注，500亿估值背后，是资本对赛道远期前景依旧抱有期待，可资本也开始学会了筛选。
本文由 AI宝库自动抓取并翻译，版权归原作者所有；如需查看原文请点下方链接。
* Reissued for technical reasons.
The time when the value of the repurchase income had risen, when only a prototype, a concept story, had been able to obtain high valuations, had passed. It is very difficult to get out of the real business ring as the main battleground, and the head enterprise needs to look for a new path, which can go even earlier than a new energy car.
Ji-won is too hard to fight.
On 21 October, Jiwon announced seven consecutive working days of publish, starting on 22 October, covering data sets, operating platforms, reloading robot G2Max, Genee Master with the Agent Framework, Simulation Platform, base LLM and group experience learning paradigms.
This is a broad context, and the attitude of capital markets towards being smart turned a 180-degree turn at the beginning and end of the year. While the whole smart track was squeezing into the capital market in the early stages, most of the IPOs became a mirage.
In particular, as a global front player, Jiwon was reported to have recently launched a new round of financing, with a market value of about $50 billion, but IPO has been suspended.
One detail is that the performance of human robots in the secondary market is a bit ludicrous, making capital from fanaticism to patience. The door no longer opens the green light on the concept story. Before. The market value after the launch of the Okau Tree Landing Centre is close to the point where the first day of the sale, the high-priced purchase of shares in the port of Merkamand, is now on the horizon.
However, a moratorium on IPOs does not amount to a waiver of capitalization. This is more like the choice of a one-time review: at the nodal point of the industry valuation bubble, rather than rushing to the market magnifying glass, it is better to back up the basics of the business in the track. The reason for this is that intelligent capital logic has undergone a fundamental shift, with market-judgement rulers shifting from prototypes, concepts, stories to quality of orders, repurchases by customers and real commercialization.
Selling a real machine is worth more than a penny of stock.
You think he's smart when you're in the capital market?
The Luk Lui Business Review noted that a number of people in the industry had leaked to the outside world that the market had repeatedly heard of the slow pace of his travel to Hong Kong and that if the new round of financing as reported in the Book of the Book of Finance were to be completed, he would not be able to complete the IPO this year.
This is not just a matter of intellectuals; it is a microcosm of the entire industrial development cycle. In the past, “robust” was a typical representation of a smart track for nearly two years, and there was a general consensus in the human robot industry that if human robots were labeled, a mass-marketed capital feast in the chip, semiconductor industry could be replicated.
Just because it's spring and late, it's a bit sudden that the first-tier market is willing to pay a high premium for the distant industry, with a prototype publish, a cool demonstration video, is enough to sustain the corporate valuation narrative, an early bubble.
The development of the human robotics industry has so far been tempered, in some cases by two secondary market developments.
In August, the first piece of human-form A robot technology landed on the board, with a market value of over $440 billion at a time, with a short number of trading days close to the value of the dollar, evaporating over 200 billion. In the Hong Kong stock market, the robotic parts manufacturer, Mercamand, received about 3,335 times the public oversubscription, updated the Hong Kong stock robot track bookings, and the market went out directly on its first day, and the subsequent stock prices continued to weaken.
Successive movements are sending a clear signal that capital has shifted from fanatical to patient capital. It is feared that second-tier market investors do not want to pay for the concept of human robots alone, and businesses have the capacity to purchase real orders, market-based repurchases, with a weight being placed at an unprecedented level.
In the case of enterprises, including Ji-won, following the rules of chapter 18C of the Hong Kong Specialized Technology, if they submit an application for listing, and if a new round of financing is actually completed before they are listed, they must wait 120 days before they can move the listing process forward, without an exchange exemption.
The new financing objectively lengthened the listing schedule of other non-IPO enterprises. However, leaving aside objective limitations at the level of rules, a change in the industry environment is an inescapable bottom-line trigger.
After all, the regulatory review threshold has been significantly raised. The release of signals from Hong Kong's offices is planned for the first half of next year, in consultation with the 18C Specialized Technology Listing System, which contains a review of the market value threshold, which will have a direct impact on the fate of dozens of furniture-smart enterprises in the queue. Similarly, in June, the Listing Regulations for Unit A Section were revised to give priority to robotics and intelligence, with a clear focus on access to the market, the financial authenticity threshold, and the risk of disease reporting.
Wind statistics indicate that all Hong Kong enterprises with more than 50 furniture-based smart tracks are waiting in line. However, as a result of a combo process, the market in the industrial chain has become a real divide: upstream parts and components manufacturers are relatively smooth on the market, and cyborg robots, fairy-wire intelligence, Mekamande and Benmong technology have completed the registration of Hong Kong shares, and the human integrator enterprise is clearly more resistant to breaking through, and it is unlikely that it will be possible in the short term to see a case of a clock-calling.
It's a long way to go.
Theoretically, the equity book is a fully open examination, with all income structures, the source of the order, the composition of the customer, all of which are spread out to investors and industry counterparts. But has there been an open dispute in the course, and the market has begun to distinguish between a single income from a market-based business customer, or where?
Against this background, companies wishing to be listed would be criticized if they were blown out of real income without a sufficient amount of money. It is clear that very few enterprises are willing to confront public torture of income authenticity from capital markets, investors and the general public when they are not adequately prepared. The suspension of IPO, with a window period to sharpen business indicators, became a realistic choice for some of the later players.
It appears that even large factories have had to travel to demand, and gradually there is an ideal belief that, in the future, operational indicators will be more important than time-triggering, and can withstand auditing and secondary market scrutiny.
Now that the window in the capital market has changed, the outside is only a source of pressure, and the real impact on how far an enterprise can go remains in its own business chassis.
Ji-won's rival, he's in business or he's in gold.
According to the pattern of competition in the human robotics sector, Ji-won and Woo-hu form the two oligarchs of the track.
In the first half of 2026, SAG ' s statistics show that 8,400 pieces of IQD and 5900 pieces of Waki ' s, both of them have taken more than 70 per cent of the world ' s market share, while the remaining small and medium-sized producers have not been able to reach a thousand in the first half of the year, and the gap between the size of their head and waist has widened.
Woo-tree had been profitable before. On the surface, Ji-won's biggest opponent is a group of friends from the same track.
But penetrating the industry's appearance, the development barriers common to the entire track are at the core of the examination on the brain. The real test comes from the gold in the business itself.
The whole industry is being examined for the real color of income. According to the Union for the Application of Human Robotic Sites, there were 218 winnings of open human robotics in the first half of 2026, totalling 1,723 million yuan. Of these, 55.5 per cent were procured by educational and scientific institutions, 20.6 per cent by the Government and the State, and only 21.1 per cent were actually ordered from industrial enterprises. In addition, most of the industrial business orders remain in prototype testing, exhibition halls, hands-on piloting and are not commercialized factory production lines.
Assuming that a large number of orders carry scientific project attributes, the budget follows the project cycle, making it difficult to generate a continuous repurchase. Once the project is completed, there is a risk of volatility in the counterpart collection, which is also the point of risk that is the focus of ongoing regulatory inquiries and secondary market investors. This is not a special case for a particular enterprise, but rather an industry reality that is shared by the vast majority of human processors.
On the positive side of the picture, for example, from 2 to 3 November, six G2 robots are expected to go into a 24-hour live operation at a real convenience store to complete the real scene validation of the whole house technology.
Also, at the 2026 Meeting of Ji-won Partners, Deng Taehua, founder and CEO of Ji-won, revealed the status of his collections since his creation: “From $300,000 in his first year of existence to more than $60 million in his second year of existence, last year (2025), he surpassed the $1 billion harvest. Twenty times the annual increase. This year ' s target will be further multiplied. In 2027, it's going to cross $10 billion."
But these do not mean that the product is fully mature. Delivery under a statistical caliber consists of a wheeled, half-sized machine, full-sized human robotic deployment and room for upgrading.
The country ' s large number of integrator enterprises, with high rates of external procurement of core meta-measurers such as brakes, servers, controllers and so forth, a high degree of convergence in the supply chain, and the ease of making walking and performing robots at the prototype level, but with more product differentiation remaining in appearance and demonstration, operational capability of real, non-structured industrial scenes, and the need for a reversal of mass situational data. Thus, even in the case of intellectuals, the equipment can be deployed in increasing quantities, but from demonstration to more stable work, there is no small distance.
The company CEO Duntawa had stated that “no rush to net profit”, and he also stressed that Jiwon's current strategy was to maintain saturated investment in R & D to build product competitiveness.
Under the redevelopment track, there is an objective pressure on the cash flow of the enterprise. Hardware iterative, algorithmic training, team manpower, market expansion, each of which continues to consume funds. Ji-won has not yet made a profit and is dependent on external financing.
In retrospect, competition for export and market share among peers is only a competition at the surface level. The order structure is optimized, the physical landing capacity of the product, and the path of loss recovery, and these are the propositions at the level of the business chassis, which is the internal level that intellectuals need to cross.
Compared to the increasing spree of front-line players in the country, the Rudd Business Review has observed another trend in which it is difficult to get out of a real business closed ring as the main battleground, and head-run firms need to look for new ways to break through, which can even go earlier than new energy vehicles.
Why don't we go out ahead of each other on the inside? Hey.
It is not unreasonable to assume that there will be an accelerated outburst.
Looking back at the path to the development of the new energy automobile industry, most domestic producers have experienced the development sequence of first deep-growing local markets, accumulating technology and scale, and then going out to sea. In the case of smart races, there are different possibilities in the industry: domestic-born robots have supply chains, cost advantages, and access to the sea can become an important engine of growth earlier.
As far as big data are concerned, the data show that during the first quarter of this year, Chinese human robot exports increased by 210 per cent each year. According to the 14 October magazine of Bloomberg, China is expected to lead a $5 trillion human robot market.
With regard to Ji-won specifically, the joint president, Kang Qinghong, introduced the company ' s planned overseas sales share of over 30 per cent by 2026, which is expected to exceed 50 per cent in the future.
The above data mean that grinding the basics of the business to the inside, taking out the overseas incremental market, can be a more realistic way out for head-to-head complexes, with an opportunity to be a critical second curve in this solution.
The Chinese robotics have a strong advantage in going out to sea, and Morgan Stanley has made a judgement that the human robotics industry is developing at a level equivalent to a new energy car from 2018 to 2020, with a 100-fold increase in global exports over the next five years, with Chinese manufacturers taking the lead in the industrial chain. New energy vehicles rely on energy-replacement logic to open overseas markets, human robots rely on labour to replace new demands, market space, speed of expansion, industrial chains of speech are all the better, and the wave surge is stronger.
The disadvantage lies in whether external markets will accept access to products on a scale that tests the ability to comply with products, including the resource person, while the first wall from competitive markets is not easily overlooked.
According to reliable sources, with Tesla revealing that his third-generation human-form robot Optimus (Optimus) is ready for large-scale production, the eyes of the market again focus on this enterprise, just as Tesla flew in the previous period, the domestic-form robot brand cannot ignore the competitive pressures from Tesla on the global market. Mask would even say that Optimus human robots and Cybercab would make up the bulk of Tesla's value.
The second wall is that the path to industrialization is not an easy one and that hardware reliability, physical AI algorithmization capability, cost control, industry compliance standards and data closure construction remain central constraints to the development of the intellectuals and others robotics industry.
Looking back, there is a long industrial trail between eye-port queues, technical demonstrations and a sustainable business. Instead of continuing to fight on the inside of the domestic stock track, it is better to spend resources on grinding products and on the global market for real commercial customers from overseas markets, but this road will be more difficult.
The suspension of IPO by the explosion of Ji-won is not an industry turning point, and the first-level market is still willing to bet on long-term industrial space for human robots, behind a $50 billion valuation, because capital still looks forward to the future of the track, but capital is beginning to learn to filter.
Auto-fetched and translated by AI Gem Box. All rights belong to the original author.