类人机器人，这种由软件驱动的双足机器，如今已成为人工智能最具体的成果之一。它们看起来像人、像人一样移动，并且越来越能像人一样学习，使其成为面向工厂、仓库，最终还有家庭的潜在通用自动化平台。与传统固定且专门化的工业机器人不同，理论上，一台类人机器人可执行多种任务。经过一个多世纪关于其即将到来的科幻想像之后，它们现在已经开始真正进入实际应用轨道。
目前的竞争常被界定为中国对美国。中国凭借政府支持、工业政策及深厚的制造业基础，具有早期动能；Morgan Stanley 指出，这种支持正在帮助中国在当前发展阶段取得领先。美国则依赖其现有的 AI 领先地位，以及像 Nvidia 和 Tesla 这样的公司推进激进的时间表进行回应。Barclays 估计目前的类人机器人市场仅有约 20 到 30 亿美元；但预测差异很大，最高情景到 2035 年可达 400 亿美元，更乐观情境高达 2000 亿美元，且增益会从 Big Tech 扩散到致动器制造商、自动化企业和精密工程公司。
从更长期来看，Morgan Stanley 在纳入供应链、维修和服务后，预估到 2050 年市场规模有可能超过 5 兆美元。本文节目中，Bloomberg Primer 介绍了 Nvidia 企图提供类人机器人的“大脑”算力，训练在实体环境中工作的 AI 模型；也提到 Elon Musk 将 Optimus 推广到大众市场，以及 1X 在矽谷采取较为温和的家庭导向规划。近期最清晰的应用场景是制造业：工厂劳动力短缺，再加上 AI 模型对物理现象的理解能力增强和实时自适应能力，正在把类人机器人从实验室推向生产线；已有 Tesla、Agility Robotics 等公司进行小规模测试。当前热度与落地现实仍有差距，但这种差距正在缩小。
Humanoid robots, or bipedal machines powered by software, are becoming one of the most tangible outputs of the AI revolution. They look like us, move like us, and increasingly can learn like us, which makes them appealing as potentially general-purpose automation platforms for factories, warehouses, and eventually homes. Unlike traditional industrial robots that are fixed and specialized, one humanoid could perform many tasks. After more than a century of science-fiction expectations of their imminent arrival, they are now beginning to gain real momentum.
The race is now often framed as China versus the United States. China has early momentum from government support, industrial policy, and a deep manufacturing base; Morgan Stanley says this backing is already helping China lead at the current stage of development. The U.S. counters with current AI leadership and firms such as Nvidia and Tesla pushing aggressive timelines. Today, Barclays estimates the humanoid market at only about $2 to $3 billion, while forecasts diverge sharply: up to $40 billion by 2035 and as high as $200 billion in more optimistic scenarios, with value spreading beyond Big Tech to actuator makers, automation firms, and precision engineering companies.
Longer-term, Morgan Stanley estimates that including supply chains, maintenance, and services, the market could exceed $5 trillion by 2050. The episode focuses on Nvidia’s aim to supply the “brains” for humanoids and train AI models for physical work; Elon Musk’s goal of taking Optimus to the mass market; and 1X’s more measured plans for home use. Manufacturing is the clearest near-term use case: factory labor shortages, plus AI models with better physical understanding and real-time adaptation, are moving humanoids from labs to production lines, with pilots by Tesla, Agility Robotics, and others. Current hype still exceeds deployment at scale, but the gap between promise and real-world implementation is narrowing.