True

过去一年，一位长期做科技产业商业化的受访者，向几乎所有能坐下来谈的具身智能公司反复问同一个问题，你们到底怎么赚钱。他这么问，是因为自己就是做商业化出身的，客户正是这些做本体、做方案、做模型的公司。虎嗅这篇具身智能商业化调查给出的答案，落在标题上，量产的元年不等于商业化的元年，行业的钱，大多还在体系里自己转圈。

Over the past year, a practitioner long involved in tech-industry commercialization kept asking the same question to every embodied-AI company he could sit down with: how do you actually make money? He asks because commercialization is his trade and his clients are precisely these makers of bodies, solutions and models. Huxiu's investigation lands the answer in its headline: the year-one of mass production is not the year-one of commercialization, and much of the industry's money is still circling inside the system itself.

数据端先把热度摆出来。沙利文的口径里，中国具身智能机器人市场规模从2023年的2.9亿元升到2025年的28.5亿元，复合增长率216.1%，预计到2030年达到684.5亿元。人形机器人这边，高工机器人产业研究所的数据是2025年国内出货1.8万台，同比增超650%，2026年有望提到6.25万台，业内乐观估计能到10万到20万台级。

The data lays out the heat first. By Sullivan's measure, China's embodied-intelligence robot market grew from 290 million yuan in 2023 to 2.85 billion in 2025, a 216.1% compound rate, and is projected to reach 68.45 billion by 2030. For humanoids, GGII puts 2025 domestic shipments at 18,000 units, up more than 650% year on year, with 2026 expected to reach 62,500, and optimists in the industry sizing it at 100,000 to 200,000.

成本是另一头。人形机器人的单机成本，已经从百万元级一路压到几十万元。成本下探是量产能起来的前置条件，但它只是必要条件，不是充分条件，便宜了不等于有人买单。

Cost is the other end. Per-unit cost for humanoids has come down from the million-yuan tier to the tens-of-thousands. Falling cost is a precondition for scale, but only a necessary one, not a sufficient one. Cheaper is not the same as someone willing to pay.

IT桔子的统计更能看出结构。2025年具身智能领域共发生329起融资，金额398.9亿元，同比翻了三倍。关键是集中度高，单轮金额最高的9起事件，就拿走了超过100亿元。银河通用12月拿到3亿美元的B+轮，乐聚拿了15亿元Pre-IPO，星动纪元10亿元A+轮。关键零部件这一层交易最密，131起占了40%，本体层只有62起。

IT Juzi's tally shows the structure. In 2025 the sector saw 329 funding events worth 39.89 billion yuan, triple the year before. Concentration is the striking part: the nine largest single rounds together took over 10 billion yuan. Galbot closed a 300-million-dollar Series B+ in December, Leju raised a 1.5-billion Pre-IPO, and Xingdongjiji took a 1-billion Series A+. The components layer had the most deals, 131, or 40%, while the body layer had only 62.

循环的特征就藏在这里。资本以产业资本CVC为核心，百度、阿里、美团、京东这些大厂通过直接投资或战略协作锁定头部一体化厂商，站队效应把资源越拢越紧。可资金大多还在体系内部打转，谁投谁、谁用谁，终端市场能不能真的造出血来，还没有被充分证明。

The circularity hides right here. Capital is anchored on corporate venture arms, and giants such as Baidu, Alibaba, Meituan and JD lock onto the leading integrated players through direct or strategic investment; the alignment effect tightens resources ever further. Yet most of that money still turns inside the system, one investor into one user, and whether the terminal market can genuinely produce cash has not been adequately proven.

判断落到时间点。奕丰基金的投资经理李净说，业界把2026年定成量产元年，但量产不等于商业化落地，从样机展示到企业主掏钱买单，中间隔着成本与实用性的鸿沟，现实的规模商业化节点可能出现在2027到2028年。均普这类已经开工的产线也给了审慎的注脚，签了约3500万订单，但今年产量规模，应该不会过万台。产线在转，产量能不能变成订单和利润，才是这场大考真正的考题。

Judgments pin down the timing. Li Jing, a fund manager at Lifeng, says 2026 is being named the mass-production year, but production is not deployment. Between a sample on display and a company owner actually paying, cost and practicality stand in the way, and the realistic point of scaled commercialization may land in 2027 to 2028. A line already running, such as Junpu's, footnotes the caution, roughly 35 million yuan in signed orders, yet this year's output, probably not past ten thousand units. The lines are turning; whether that output converts into orders and profit is the real question on the exam.