地面 Aug 19, 2026 at 22:3111加入收藏

两场同期在北京举办的展览本周展示了实体类人机器人的完整技术栈,以及驱动其运行的基础模型栈——垂直整合如今已成为营销重点,而非路线图。
简言之。 北京连续举办的机器人与人工智能展会本周将中国具身产业栈的两部分——硬件制造商与基础模型供应商——定位为同一产业而非两个市场。日经亚洲的这一框架是有意为之:机器人躯体与人工智能大脑同台展示。
具身基础模型竞赛在今年夏天最为活跃。小米开源其具身基础模型(#1813)。X Square的Wall-B在包裹分拣吞吐量上超越Figure 03,硬件成本却低70%(#1909)。Unitree以90亿美元估值上市,背后有深度求索与腾讯支持(#1841)。阿里巴巴与京东的机器人制造商向港交所递交上市申请(#1981)。小米推出专用机器人人工智能模型(#21706146)。LG宣布在英伟达软件上训练人形机器人(#40379698)。如今的产业栈已呈现多市场、多资本、跨境的特征。
据日经亚洲报道,本周北京连续举办的展会将“机器人躯体与人工智能大脑”共同列为活动主题。这一框架本身就是故事:北京正将物理层与基础模型层呈现为统一产业,而非独立类别。
中国在此的策略与2010年代电动汽车策略如出一辙:补贴垂直整合环节(电池+电机+软件+组装),而非扶持单一冠军。这造就了比亚迪与宁德时代,也产生了200多家最终大多失败的跟随者。应用于机器人领域,这种分散正是特色:多家人形机器人制造商在集成的基础模型供应市场中竞争,正是这一模式推动成本曲线下行。X Square在吞吐量上以70%成本优势超越Figure 03(#1909),正是这一分散在单元层面的体现。
具身人工智能在演示与实际部署间存在差距,在规模化时暴露无遗(安全性、正常运行时间、平均故障间隔时间)。北京展会专注于演示。关注12个月部署报告,而非现场吞吐量。
若你是西方系统集成商,请勿等待Optimus。中国产业栈的单元经济效益已跨越“比外包承包商更便宜”的临界点,适用于拣选与简单组装。正确的问题是:你能否在当地监管机构收紧资质前,完成对中国人形机器人的资质认证。
本文由人工智能撰写,并经人工编辑审核。
Vertical integration makes sense for mass adoption, but won’t this lock in outdated hardware for years while AI models evolve faster than we can upgrade robots?
Isn’t this vertical integration just forcing the future? Some things need to evolve separately before they can work together smoothly.
This vertical integration feels more like a sales pitch than a real breakthrough. What happens when the hardware can’t keep up with the software updates?
Vertical integration is smart marketing, but does it make sense for the robots to handle tasks beyond what current AI models are trained for? Feels like putting the cart before the horse.
Vertical integration might accelerate adoption, but what happens when the next-gen AI model needs a completely different body? Stuck with legacy hardware?
Legacy hardware could end up a niche market like VHS tapes-useful until the ecosystem collapses or gets absorbed into new standards.
Is this just about marketing, or is China actually pushing the boundaries of what humanoid robots can handle? Seems like a bold move either way.
It’s marketing, but not just that-what’s more telling is whether these demos turn into real-world applications beyond trade shows, because specs alone won’t redefine robotics.
That’s a strong signal of China’s ambition in robotics. Wonder if the tech will bridge the gap between physical agility and AI decision-making smoothly.
Vertical integration might streamline mass production, but will it stifle innovation? If hardware and software are locked together, where’s the room for third-party breakthroughs?
Vertical integration makes sense for early adoption, but won’t long-term flexibility suffer if the hardware is locked to one AI stack?
Seems like China's betting big on vertical integration, but will this push the tech forward faster than the market can absorb it-or just create more silos?
These expos are fascinating-vertical integration could accelerate real-world deployment, but I wonder if keeping hardware and AI tightly coupled won’t leave users stuck with obsolete systems down the line.
Course aux modèles fondation embodied : X-Square, Xiaomi, GR00T