According to a post on the DaXiao Robotics WeChat official account, this round of financing attracted participation from institutions such as Fortune Capital, Shenzhen Capital Group, Shanghai Sci-Tech Innovation Fund, MetaX, Sunic Capital, Fosun RZ Capital, Huakong Fund, Lingang Special Area Fund, and Yuzi Zhangquan, with existing shareholder SenseTime's Guoxiang Capital continuing to increase its investment. As of now, DaXiao Robotics' cumulative financing in the first half of 2026 has reached several hundred million US dollars.
This is a star financing round, but it's not just a story of hot money in the embodied intelligence track. Viewed within the context of SenseTime's business evolution, DaXiao Robotics appears more like SenseTime's renewed bet on "Physical AI" following Jueying.
This isn't the first time SenseTime has attempted to put AI into the physical world. Previously, SenseTime's Jueying unit undertook the task of entering the intelligent vehicle industry for SenseTime. Public information shows that by the end of 2025, Jueying had collaborated with over 30 domestic and international automakers, covering 188 vehicle models, with cumulative shipments nearing 5.5 million units. Unfortunately, it failed to land a "blockbuster" model.
Therefore, as a peripheral supplier, Jueying also failed to attain an industry position commensurate with SenseTime's AI capabilities.
There have been no perennial winners in China's intelligent driving industry over the past few years. In the dynamic development of the industry, automakers no longer need just algorithm suppliers; they require engineering delivery, cost control, data closed-loop, OTA iteration, and vehicle integration capabilities. Huawei has established a strong presence through its vehicle partnership model; companies like Yuanrong Qihang and Qingzhou Zhihang have strengthened market recognition through projects with leading automakers; while Horizon Robotics entered the core supply chain with chips and computing platforms. In contrast, Jueying possessed AI technology, mass production projects, and the SenseTime brand, but failed to become a key variable in automakers' high-level intelligent driving strategies.
To some extent, this also explains the logic behind SenseTime promoting its "1+X" organizational adjustment, leading to Jueying gradually moving towards independent development. In 2025, SenseTime's X Innovation Business revenue declined, partly because the intelligent driving business was no longer consolidated into the financial statements starting August of that year. For SenseTime, Jueying remains active in the intelligent vehicle race, but from the group's perspective, it no longer plays the role of a core growth engine.
Re-Entering the Fray in Embodied Intelligence
The birth of DaXiao Robotics, to some extent, coincides with the rapidly heating window period for embodied intelligence in the primary market, aptly described as being "born at the right time."
According to the disclosure, the funds from this financing round will be primarily invested in world models and integrated hardware-software commercial solutions, specifically including the Enlightenment World Model 3.0, large-scale embodied training, end-side direct-drive control, as well as commercial scenarios like smart retail, security patrols, cultural tourism, and hotels. Its goal is not a single robot body, but rather an attempt to become the "brain" behind different robot hardware.
This story is more compelling than Jueying's prospects, and also more "capital-intensive."
World models are not ordinary algorithm modules. They require the model to understand space, objects, actions, causality, and physical laws, be able to generate scenes, predict changes, and guide robots to complete long-sequence tasks. Compared to language models, world models need to process video, 3D space, sensor data, motion data, and simulation environments; compared to intelligent driving, embodied intelligence must deal with more unstructured scenarios and more complex interactive objects. Training such a system essentially entails a long-term consumption of computing power, data, engineering, and capital.
The same path is being taken overseas. NVIDIA's Cosmos regards world models as the infrastructure for Physical AI, used in scenarios like robotics, autonomous driving, and industrial vision. However, this route has an extremely high threshold. Industry research shows that the training costs for cutting-edge AI models are still rising rapidly. World models are not only expensive to train, but inference and deployment costs could also become bottlenecks for commercialization.
This is no small issue for DaXiao Robotics. DaXiao Robotics has raised several hundred million US dollars cumulatively this year, which may seem substantial. However, for world model R&D, real-world scenario data collection, simulation platform construction, end-side deployment, robot control, and industry delivery, this funding may not be ample. Especially as the company's goal shifts from model demonstrations to commercial-grade delivery, capital consumption will extend from the training side to hardware adaptation, scenario pilots, operation and maintenance systems, and channel building.
A more realistic problem is that SenseTime is not a financially agile parent company.
From 2018 to 2024, SenseTime accumulated losses exceeding 54 billion yuan. In 2025, SenseTime's revenue grew to over 5 billion yuan, and its net loss narrowed significantly, but the full-year loss was still 1.782 billion yuan. Generative AI has become SenseTime's new revenue mainstay, and the company is also reducing losses, but this doesn't mean it has escaped the model of long-term investment for growth. Large models, AIDC, intelligent agents, visual AI, and X Innovation Business all require funds. SenseTime does not have much surplus capacity to sustain a new story that only burns money and fails to achieve closure for a long time.
This adds another layer of practical meaning to DaXiao Robotics' financing. On the surface, it's a star financing round in the embodied intelligence track; observed within the SenseTime system, it appears more like an external financing arrangement for a high-cost innovation business. World models, end-side control, and scenario delivery all require continuous investment, while SenseTime itself is still in a cycle of loss reduction. For DaXiao Robotics to continue advancing the Physical AI narrative, it cannot rely solely on SenseTime's infusion; external capital must also share the burden of the next stage's R&D and commercialization costs.
Another variable associated with DaXiao Robotics is Wang Xiaogang.
A distinct feature of the current embodied intelligence industry is the rapid emergence of young, technically-oriented founders. Wang Xingxing of Unitree Robotics, Peng Zhihui of Zhiyuan Robot, Chen Jianyu of StarMove, Wang He of Galaxy General, and Jiang Zheyuan of Songyan Power represent different paths taken by post-90s and post-95s entrepreneurs in humanoid robot bodies, motion control, embodied large models, and integrated hardware-software systems. Most have strong technical backgrounds, rapid financing rhythms, frequent product releases, and are more adept at creating industry buzz through demos and engineering iterations.
In contrast, SenseTime co-founder, executive director, and chairman of DaXiao Robotics, Wang Xiaogang, is not the typical young entrepreneur in this wave of embodied intelligence. He has experienced the cycles of visual AI, smart cities, intelligent driving, and large model commercialization, long operating at the intersection of AI research, industrialization, and corporate governance. This background might bring DaXiao Robotics stronger technical organization capabilities and industrial resources, but it may also naturally impart the inertia of a large SenseTime-like organization to the company.
This is precisely the most "special" aspect differentiating DaXiao Robotics from other star projects.
Embodied intelligence is still in its early stages, with technical routes, commercial scenarios, cost structures, and industry divisions of labor yet to be defined. The advantage of young teams lies in being lightweight, fast, and daring to experiment; DaXiao Robotics' advantage lies in SenseTime's technological, capital, and resource ecosystem. However, SenseTime's experience over the past few years has proven that technological leadership does not automatically translate to commercial leadership. Smart cities and intelligent driving have both been constrained by objective factors; large models face pressure from internet platforms and cloud vendors; embodied intelligence will similarly not be ranked solely based on papers, evaluations, and financing amounts.
Jueying serves as a ready reference. SenseTime could enter the intelligent vehicle industry with its established capabilities but failed to secure a high position in the intelligent driving industry chain. DaXiao Robotics must solve this same problem, needing to rapidly integrate world models, environmental data, end-side control, and industry scenarios to form stable revenue and replicable delivery, seizing every advantage.
Capital has given DaXiao Robotics a high starting point, but what SenseTime truly needs is not another financing showcase, but a new business capable of realizing the imagination of AI industrialization. The industrial position Jueying failed to complete, DaXiao Robotics must strive for anew. Only this time, the battlefield is earlier, the story grander, and the money burns faster.
This article is from WeChat Official Account: Economic Observer Perception , Author: Wu Yuan
