中美具身智能头部企业发展路线差异显著。美国企业侧重深研核心技术，依托先发生态与充裕资本走长期深耕路径，我国企业侧重规模落地，依托供应链效率与成本优势走快速商业化路径。这种分野源于两国产业禀赋差异，各有其合理性。我国应用层已形成先发优势，但生态完整度仍有欠缺。
Chinese and American embodied intelligence leading companies have significantly different development paths. American companies focus on deep core technology research, relying on first-mover ecosystems and ample capital to take a long-term deep cultivation path. Chinese companies focus on scaled deployment, relying on supply chain efficiency and cost advantages to take a rapid commercialization path. This divergence stems from differences in the two countries' industrial endowments, and each has its rationality. China has formed first-mover advantages at the application layer, but ecosystem completeness still has shortcomings.
发展重心上，美国偏技术壁垒构建，我国偏应用驱动与规模铺开。美国头部企业更注重技术壁垒构建和单场景深度验证。Figure AI在宝马工厂历经11个月部署周期，第10个月才实现装配线规模化运行。波士顿动力数十年深耕运动控制，Atlas代表全球动态运动能力最高水平，但商业化进程缓慢。谷歌通过Gemini多模态大模型和DeepMind基础研究为行业提供AI能力，不直接参与本体制造竞争。我国企业更强调规模量产和多场景快速铺开。赛迪研究院数据显示，2025年全球人形机器人出货量约1.7万台，我国企业占比超80%。宇树以32.4%的全球出货份额位居第一，业务覆盖60余个国家和地区，应用场景覆盖工业制造、智慧巡检、应急救援等多个领域。
On development focus, the US leans toward building technical barriers, China toward application-driven and scaled expansion. US leading companies emphasize technical barrier building and deep single-scenario validation. Figure AI went through an 11-month deployment cycle at BMW's factory, only achieving scaled assembly line operation in the 10th month. Boston Dynamics has深耕 motion control for decades; Atlas represents the world's highest level of dynamic motion capability, but commercialization is slow. Google provides AI capabilities to the industry through Gemini multimodal large models and DeepMind basic research, not directly competing in body manufacturing. Chinese companies emphasize scaled mass production and rapid multi-scenario expansion. CCID Research Institute data shows global humanoid robot shipments in 2025 were about 17,000 units, with Chinese companies accounting for over 80%. Unitree ranks first with 32.4% global shipment share, business covering over 60 countries and regions, with application scenarios covering industrial manufacturing, smart inspection, emergency rescue, and more.
资本支撑上，美国偏高估值长期投入，我国偏快速出货商业验证。美国头部企业以高估值支撑长期研发投入。据报道，2026年1月，Skild AI完成14亿美元C轮融资，估值超140亿美元，专注通用机器人基础模型研发，7个月内估值翻三倍。2026年3月，Physical Intelligence据报道以超110亿美元估值筹集约10亿美元，同样专注基础模型研发，在无产品交付情况下获得高估值融资。我国企业以快速出货验证商业模式。2026年上半年国内具身智能融资322笔、超900亿元，但资金高度集中于整机企业和零部件领域，基础研发获投占比偏低。融资评价标准偏重出货量、订单金额和场景覆盖率，基础研发的长期价值难以在短期财务指标中体现。
On capital support, the US leans toward high valuation long-term investment, China toward rapid shipment commercial validation. US leading companies support long-term R&D investment with high valuations. Reportedly, in January 2026, Skild AI completed a $1.4 billion Series C at a valuation exceeding $14 billion, focusing on general robot foundation model R&D, with valuation tripling within 7 months. In March 2026, Physical Intelligence reportedly raised about $1 billion at a valuation exceeding $11 billion, also focusing on foundation model R&D, obtaining high-valuation funding without product delivery. Chinese companies validate business models through rapid shipment. In the first half of 2026, domestic embodied intelligence financing totaled 322 deals exceeding 90 billion yuan, but funds were highly concentrated in complete machine enterprises and components, with basic R&D investment share low. Financing evaluation criteria emphasize shipment volume, order amounts, and scenario coverage; basic R&D's long-term value is hard to reflect in short-term financial metrics.
路线差异根植于产业禀赋与技术积累路径。美国AI基础层先发优势支撑技术壁垒构建路线。斯坦福大学《2026年AI指数报告》显示，中美AI模型性能差距已缩小至2.7%，基础模型层面并不构成主要差异，差距主要体现在AI基础设施和全球开发者生态层。我国制造和场景优势导向应用驱动与规模铺开路线。我国制造业体量和供应链体系为企业提供了从硬件制造到场景落地的完整条件，企业无需跨越高技术壁垒即可实现商业闭环，天然导向应用驱动与规模铺开路线。美国制造业占比较低、工业场景开放度有限，企业难以依托供应链和场景优势快速变现，更倾向于通过构建技术壁垒获取竞争优势。
The path divergence is rooted in industrial endowments and technology accumulation paths. US first-mover advantages in AI foundational layers support the technical barrier building route. Stanford University's "2026 AI Index Report" shows the China-US AI model performance gap has narrowed to 2.7%; at the foundation model level there is no major difference—the gap mainly manifests in AI infrastructure and global developer ecosystem layers. China's manufacturing and scenario advantages guide the application-driven and scaled expansion route. China's manufacturing scale and supply chain system provide companies complete conditions from hardware manufacturing to scenario deployment; companies can achieve commercial closed loops without crossing high technical barriers, naturally导向 application-driven and scaled expansion routes. The US has a lower manufacturing share and limited industrial scenario openness; companies struggle to quickly monetize through supply chain and scenario advantages, tending more toward building technical barriers to gain competitive advantage.
资本偏好差异强化了路线差异。美国风险资本的传统偏好和退出机制天然适配长期基础研发投入，与技术壁垒构建路线形成正反馈。我国产业资本和政府引导基金占比高，资本结构天然偏好短期可验证的商业成果，与应用驱动路线形成正反馈。资本偏好与路线选择相互强化，使两国路线差异进一步固化。应用驱动与规模铺开是我国在禀赋约束下的合理起步，当前已取得显著成效。但产业生态的完整性不仅取决于应用繁荣，还需要技术壁垒构建能力、AI基础设施与开发者生态、长期研发资本等系统条件的支撑。
Capital preference differences reinforce path differences. US venture capital's traditional preferences and exit mechanisms naturally fit long-term basic R&D investment, forming positive feedback with the technical barrier building route. China's industrial capital and government-guided funds have high proportions, with capital structures naturally preferring short-term verifiable commercial results, forming positive feedback with the application-driven route. Capital preferences and route choices mutually reinforce, further solidifying the two countries' path differences. Application-driven and scaled expansion is China's reasonable starting point under endowment constraints, and has achieved notable results. But industrial ecosystem completeness depends not only on application prosperity but also on technical barrier building capability, AI infrastructure and developer ecosystem, long-term R&D capital, and other systemic conditions.
我国亟须完善举措，补齐上述短板，加速推动产业生态从应用起步向系统成熟跃迁。加强技术壁垒构建能力，推动应用层成果向技术能力转化。依托十五五规划关于前瞻布局未来产业的要求，设立具身智能基础研究专项，对通用模型训练、仿真平台等基础研发给予长期稳定投入支持。提升AI基础设施与开发者生态，补齐产业生态关键短板。推动已建成训练场和数据平台从封闭运营转向开放共享，推进场景运营数据标准化，使其具备跨场景、跨企业的复用价值。优化资本结构，为基础研发提供长期稳定支撑。探索对技术壁垒构建型企业实行能力估值与营收估值并重的评价体系，鼓励保险资金、养老基金等长周期资本参与具身智能基础研发投资。
China urgently needs to improve measures, fill the above shortcomings, and accelerate the industrial ecosystem's transition from application start to systemic maturity. Strengthen technical barrier building capability, promoting application-layer results into technical capability. Relying on the 15th Five-Year Plan's requirements for forward-looking future industry布局, establish embodied intelligence basic research special projects, providing long-term stable investment support for general model training, simulation platforms, and other basic R&D. Improve AI infrastructure and developer ecosystem, filling key ecosystem shortcomings. Push built training grounds and data platforms from closed operation to open sharing, advancing scenario operation data standardization to give them cross-scenario, cross-enterprise reuse value. Optimize capital structure, providing long-term stable support for basic R&D. Explore implementing an evaluation system balancing "capability valuation" and "revenue valuation" for technical barrier building enterprises, encouraging insurance funds, pension funds, and other long-cycle capital to participate in embodied intelligence basic R&D investment.
