摩根士丹利发布人形机器人月度跟踪报告《人形机器人视野：让机器人投入工作》。报告指出，行业竞争重心正从供给端转向需求端，核心问题变成"谁来雇佣机器人"。落地方面，宝马将人形机器人投入电池与零部件生产，现代围绕物流和零件排序搭建Atlas训练环境，丰田在考虑最高需40万台机器人的自动化计划，韩国、新加坡设立专用工业与公共安全场景。可靠性门槛持续抬升：小米在定义工厂任务上实现98%成功率，Figure在未见过的家庭环境中整任务成功率56%；经济可行性取决于成功率、循环时间与干预频率。公开市场开始量化人形机器人经济性，Agility的S-4披露收入与经营亏损，宇树科技上市后估值聚焦ASP与竞争强度（摩根士丹利中国工业团队首次覆盖宇树科技，给予"持有"评级、目标价500元人民币，基于10年DCF，隐含42倍2027年预期市销率）。
After years of proving they can build robots, humanoid companies now face a new question: who will hire the robots? Early employers want success rates and operating hours, and we think the firms that win the first jobs will gather the data that drives the next model improvements. On the commercial side, BMW is putting humanoids into battery and component production and Hyundai is building an Atlas training environment around logistics and parts sequencing. Toyota is contemplating a broader automation program that would require up to 400,000 robots, while South Korea and Singapore are creating dedicated industrial and public-safety settings. The bar is moving toward reliability: Xiaomi reported a 98% success rate on a defined factory task, and Figure reported 56% whole-task success across unseen homes. Public markets are starting to quantify humanoid economics, with Agility's S-4 and Unitree's listing valuation. Robotics policy is both industrial policy and national security policy. The