1.91万台。去年同期是5100台。
2026年上半年，全球人形机器人出货量约1.91万台，同比增长275%以上。这个数字放在两年前，是想都不敢想的——2024年全年，全球人形机器人出货量还不到1万台。
增长是爆炸式的。但更值得关注的不是增长速度，而是增长的主力是谁。
国产厂商占据了全球出货份额的超过97%。2025年全年，这个数字还是84.7%。一年时间，从绝对优势变成了几乎垄断。
具体到公司，智元机器人上半年出货8400台，全球占比44%，超越宇树科技登顶全球第一。宇树以5900台、31%的份额紧随其后。前两名加在一起，就占了全球出货的75%。
这张榜单上，你几乎看不到外国公司的名字。特斯拉Optimus、Figure、Boston Dynamics——这些名字在媒体上很响亮，但在出货量榜单上，它们连前十都未必进得去。
从跟跑到领跑，国产怎么做到的？
两年前，人形机器人的叙事还是"美国领先、中国追赶"。特斯拉Optimus发布会刷屏的时候，国内的人形机器人公司还在做原型机。
怎么突然就反超了？
答案不复杂，三个关键词：供应链、成本、场景。
第一是供应链优势。中国是全球最大的制造业国家，机器人需要的电机、减速器、传感器、控制器——这些核心零部件，中国都有完整的供应链体系。而且不只是能做，成本还低。同样一台人形机器人，在中国生产的成本可能只有美国的三分之一甚至更低。
第二是成本优势带来的价格竞争力。当美国公司还在卖几十万美元一台的人形机器人时，中国公司已经把价格打到了几万、十几万美元。价格低一个数量级，能落地的场景就多一个数量级。
第三是场景落地速度快。中国有全世界最丰富的工业场景——汽车工厂、3C产线、物流仓库、零售门店……人形机器人从原型机到落地测试，在中国可以找到大量愿意尝鲜的客户。数据显示，上半年真实场景部署的比例已经超过70%，而去年同期只有50%。
"美国人形机器人公司还在拍发布会视频的时候，中国公司已经把机器人送进工厂干活了。路线不一样，结果就不一样。"—— 一位机器人行业资深人士
但反超的故事不能说得太满。有一个事实需要清醒面对：出货量领先不等于技术领先。目前出货的大部分人形机器人，执行的还是相对简单的重复性任务——搬运、分拣、简单操作。真正复杂的通用操作能力，各家都还在早期阶段。
美国公司的出货量少，很大程度上是因为它们选择了更难的路线——追求更强的通用能力，而不是更快的商业化落地。两条路线谁最终能赢，现在还不好说。
繁荣背后的隐忧：别让产能跑在技术前面
出货量暴涨是好事，但也有隐忧。
最大的担忧是：产能扩张的速度，可能已经超过了技术迭代的速度。
看看数据：上半年出货1.91万台，SAG机构预估全年接近6万台。很多公司都在扩产，都在抢产能，都在比谁能更快把机器人做出来。但机器人的核心能力——运动控制、手眼协调、自主决策——这些技术的进步速度，能跟上产能扩张的速度吗？
如果技术跟不上，出货量越大，问题可能越大。客户买了机器人发现不好用、干不了活、故障率高，那损失的不只是一个订单，而是整个行业的信誉。
历史上有过前车之鉴。前几年的服务机器人、扫地机器人、陪伴机器人赛道，都经历过"出货量暴涨→体验跟不上→口碑崩塌→行业降温"的循环。人形机器人会不会重蹈覆辙？
另一个隐忧是同质化竞争。国内人形机器人公司数量很多，但很多都是在做类似的东西——差不多的外观、差不多的功能、差不多的应用场景。大家都在拼价格、拼出货量，但真正有核心技术壁垒的公司不多。
好消息是，头部公司已经在往更深的技术方向走了。智元、宇树这些公司，不只是在做机器人本体，也在做核心零部件、做AI模型、做整体解决方案。它们的目标不是做最多的机器人，而是做最好的机器人。
人形机器人这个赛道，目前还在非常早期的阶段。出货量从千台到万台，只是第一步。从万台到十万台、百万台，中间需要的不是更多的工厂，而是更强的技术能力和更成熟的应用场景。
国产人形机器人拿到了上半场的领先。但下半场的比赛，才刚刚开始。
明天见。
19,100 units. Same period last year: 5,100.
In the first half of 2026, global humanoid robot shipments reached approximately 19,100 units, growing 275%+ year-over-year. Two years ago, this number would have been unimaginable - in 2024, total annual global humanoid robot shipments were under 10,000 units.
The growth is explosive. But what matters more than the speed of growth is who's driving it.
Chinese manufacturers account for over 97% of global shipments. In 2025, that figure was 84.7%. In one year, from dominant advantage to near-monopoly.
Breaking it down by company: ZhiYuan Robotics shipped 8,400 units in H1, a 44% global share, surpassing Unitree to claim the global #1 spot. Unitree follows with 5,900 units and 31% share. The top two combined account for 75% of global shipments.
On this leaderboard, you can barely find foreign company names. Tesla Optimus, Figure, Boston Dynamics - these names are loud in the media, but on the shipments leaderboard, they might not even crack the top ten.
From Follower to Leader - How Did China Pull It Off?
Two years ago, the humanoid robot narrative was "America leads, China follows." When Tesla's Optimus reveal went viral, Chinese humanoid robot companies were still building prototypes.
How did the tables turn so fast?
The answer isn't complicated - three keywords: supply chain, cost, and use cases.
First is supply chain advantage. China is the world's largest manufacturing nation. Motors, reducers, sensors, controllers - the core components robots need, China has complete supply chains for all of them. And not just capable of making them - making them cheaply. The same humanoid robot might cost a third or less to produce in China versus the US.
Second is price competitiveness from cost advantage. While American companies still sell humanoid robots for hundreds of thousands of dollars each, Chinese companies have pushed prices down to tens of thousands. One order of magnitude cheaper - and one order of magnitude more deployable scenarios.
Third is fast scenario deployment. China has the world's richest industrial scenarios - auto factories, 3C production lines, logistics warehouses, retail stores. From prototype to field testing, humanoid robots find plenty of early-adopter customers in China. Data shows real-scene deployment ratio already exceeded 70% in H1, up from 50% last year.
"While American humanoid robot companies were still shooting launch videos, Chinese companies were already putting robots to work in factories. Different routes, different results." - A Robotics Industry Veteran
But the comeback story shouldn't be oversold. One fact requires sobriety: shipment leadership doesn't equal technology leadership. Most humanoid robots being shipped today perform relatively simple, repetitive tasks - moving, sorting, basic manipulation. True, complex general manipulation capability? Everyone's still in the early stages.
US companies ship less, in large part because they've chosen a harder route - pursuing stronger general capability rather than faster commercialization. Which approach ultimately wins remains an open question.
Concerns Behind the Boom: Don't Let Capacity Outpace Technology
Skyrocketing shipments are good news, but there are concerns too.
The biggest worry: capacity expansion may already be outpacing technological iteration.
Look at the numbers: 19,100 units shipped in H1, SAG agency estimates nearly 60,000 for the full year. Many companies are expanding production, racing for capacity, competing to see who can build robots faster. But the core capabilities - motion control, hand-eye coordination, autonomous decision-making - can technology's pace match the capacity race?
If technology can't keep up, more shipments could mean more problems. Customers buy robots and find they don't work well, can't do the job, have high failure rates - and what's lost isn't just one order, it's the entire industry's credibility.
History has precedent. The service robot, robot vacuum, and companion robot tracks of recent years all went through cycles of "shipment surge - poor experience - reputation collapse - industry cooldown." Will humanoid robots repeat the pattern?
Another concern is homogenized competition. China has many humanoid robot companies, but many are building similar things - similar appearance, similar features, similar application scenarios. Everyone's competing on price and shipments, but few have real core technological moats.
The good news is that leading companies are already moving toward deeper technology. Companies like ZhiYuan and Unitree aren't just building robot bodies - they're building core components, AI models, end-to-end solutions. Their goal isn't to make the most robots, but the best robots.
The humanoid robot race is still in its very early stages. Shipments going from thousands to tens of thousands is just step one. Going from tens of thousands to hundreds of thousands to millions won't require more factories - it'll require stronger technical capability and more mature application scenarios.
Chinese humanoid robots have taken the first-half lead. But the second half is just beginning.
See you tomorrow.
