瑞财经10月14日消息，据浙江大学科技园发布，其持股孵化企业杭州宇岛人工智能科技有限公司正式宣布完成数千万元种子轮融资，投资方包括戈壁创投、浙大友创等。这家由浙江大学机器人研究院专家团队领衔研发、由浙江大学新昌联合创新中心孵化的Physical AGI公司，入孵当年即完成数千万元融资，成为浙大科技园人才项目一体化产业孵化模式的又一标志性成果。
Rui Finance reported on October 14 that, according to Zhejiang University Science Park's release, its equity-held incubated company Hangzhou Yudao Artificial Intelligence Technology Co., Ltd. officially announced completion of a tens-of-millions-yuan seed round, with investors including Gobi Partners and Zheda Youchuang. This Physical AGI company, led by an expert team from Zhejiang University's Robotics Research Institute and incubated by Zhejiang University's Xinchang Joint Innovation Center, completed tens-of-millions-yuan funding in its incubation year, becoming another landmark achievement of Zhejiang University Science Park's "talent project integration" industrial incubation model.
公开资料显示，NEXFORM新元体专注于通用具身机器人、关键部组件及具身智能应用解决方案的研发，是一家围绕通用硬件平台、开发者生态、物理智能体展开布局的Physical AGI公司。公司以AI for Physical AI为研发范式，通过AI Agentic开展Auto Research，以Physical RSI为技术路线，推动机器人在真实作业中实现自我进化。据悉，NEXFORM新元体已与多家酒店集团及物业管理公司签署合作协议，累计签约金额达数亿元，覆盖物业面积超1亿平方米，率先实现规模化数据采集、商业场景大规模落地及跨国远程作业等技术突破。
Public information shows NEXFORM focuses on R&D of general embodied robots, key components, and embodied intelligence application solutions. It is a Physical AGI company built around general hardware platforms, developer ecosystems, and physical agents. The company uses "AI for Physical AI" as its R&D paradigm, conducts Auto Research through AI Agentic, and takes Physical RSI (embodied recursive self-improvement) as its technical route, driving robots to achieve self-evolution in real operations. Reportedly, NEXFORM has signed cooperation agreements with multiple hotel groups and property management companies, with cumulative contract value reaching hundreds of millions of yuan, covering over 100 million square meters of property area, achieving technical breakthroughs including scaled data collection, large-scale commercial scenario deployment, and cross-border remote operations.
这条融资消息里，有几个点值得单独拎出来看。第一个是“Physical RSI”这个技术路线。RSI是Recursive Self-Improvement的缩写，递归自我改进，这个概念在AI领域讨论了很多年，核心思路是让系统参与自身学习过程的设计与改进。前面讨论欧拉万象时，这家家庭机器人公司也提出了Real-World RSI，逻辑是让机器人执行过程中暴露的问题转化为具体的学习需求，再据此确定需要补充什么数据、如何验证改进。NEXFORM新元体把RSI和Physical结合，提出Physical RSI，方向是一致的，都是让机器人在真实作业中自我进化，而不是依赖人工反复调试和重新编程。
Several points in this funding news deserve separate attention. First is the "Physical RSI" technical route. RSI stands for Recursive Self-Improvement, a concept discussed in AI for many years, with the core idea of letting systems participate in designing and improving their own learning processes. Earlier when discussing Ola Universe, this home robot company also proposed Real-World RSI, with the logic of converting problems exposed during robot execution into specific learning needs, then determining what data to supplement and how to verify improvements. NEXFORM combines RSI with Physical, proposing Physical RSI. The direction is consistent—both let robots self-evolve in real operations rather than relying on manual repeated debugging and reprogramming.
第二个是签约金额和覆盖面积。累计签约金额达数亿元，覆盖物业面积超1亿平方米，这两个数字放在一起，说明NEXFORM的商业模式已经有了实质性的订单支撑。它的落地场景是酒店集团和物业管理公司，这跟前面讨论的云迹科技和淘宝闪购合作做公寓外卖机器人是同一个赛道。酒店和物业属于封闭式、高密度、高频送物的场景，适合机器人替代人工完成楼内配送、巡检、清洁等任务。NEXFORM覆盖物业面积超1亿平方米，这个规模说明它的机器人网络已经进入了实质性的规模部署阶段，而不是停留在试点。
Second is the contract value and coverage area. Cumulative contract value reaching hundreds of millions of yuan, covering over 100 million square meters of property area—these two numbers together indicate NEXFORM's business model already has substantive order support. Its deployment scenarios are hotel groups and property management companies, the same track as the previously discussed Yunji Technology and Taobao Flash Purchase cooperation on apartment food delivery robots. Hotels and properties are closed, high-density, high-frequency delivery scenarios, suitable for robots replacing humans in in-building delivery, inspection, and cleaning tasks. NEXFORM's coverage of over 100 million square meters indicates its robot network has entered substantive scaled deployment, not just pilot.
第三个是浙大科技园的孵化模式。浙江大学科技园持股孵化，浙江大学机器人研究院提供技术团队，浙江大学新昌联合创新中心提供孵化支持，戈壁创投和浙大友创提供资金。这套模式把高校的技术积累、地方政府的产业资源、市场化投资机构的资金整合在一起，让一个实验室里的技术团队能在入孵当年就拿到数千万融资并签下数亿元订单。前面讨论过北京人形机器人创新中心、国网天津电力的实训场地、上海推动8000台人形机器人进工厂，这些都是地方政府和央企在推动具身智能落地。浙大科技园的模式，是高校在推动这件事，路径不同，但目标一致，都是把实验室里的技术推向真实场景。
Third is Zhejiang University Science Park's incubation model. Zhejiang University Science Park holds equity and incubates, Zhejiang University Robotics Research Institute provides the technical team, Zhejiang University Xinchang Joint Innovation Center provides incubation support, and Gobi Partners and Zheda Youchuang provide funding. This model integrates university technology accumulation, local government industrial resources, and market-oriented investment institution capital, letting a laboratory technical team secure tens of millions in funding and sign hundreds of millions in orders in its incubation year. Earlier we discussed Beijing Humanoid Robot Innovation Center, State Grid Tianjin Electric Power's training grounds, and Shanghai pushing 8,000 humanoid robots into factories—all local governments and central enterprises driving embodied intelligence deployment. Zhejiang University Science Park's model has universities driving this, a different path but the same goal—pushing laboratory technology into real scenarios.
