Insider Brief
- Chinese robotics and embodied-intelligence startups have announced a cluster of sizable funding rounds since the start of August, with capital increasingly flowing toward foundation models, world models, reinforcement learning and data systems designed to make robots more capable in real-world environments.
- Noin Intelligence raised $74 million to advance its GLOW world model and Knowin-X1 household robot, PokeBot closed a nine-figure Pre-A round to develop general-purpose embodied AI for home and service settings, and Discover Robotics raised $100 million in an Angel+ round less than a month after an initial financing exceeding $100 million.
- Financing in China’s embodied-intelligence sector reached 93.5 billion yuan, or about $13.86 billion, across 322 deals in the first half of 2026, while Unitree’s heavily oversubscribed Shanghai IPO and China’s 15th Five-Year Plan highlight investor and government support for the nation’s robotics and physical AI.
While the Unitree Robotics IPO is grabbing headlines, since the beginning of August, Chinese robotics and embodied intelligence startups have announced a cluster of sizable funding rounds around companies building the software and model layers that enable robots to perform useful work in homes and everyday environments, rather than pure hardware platforms.
Noin Intelligence
Knowin, the English-language representation of Noin Intelligence, a Shenzhen-based developer of consumer-oriented embodied AI models, announced it had closed a $74 million Angel++ round, China’s Global Times reported. The financing was led by Matrix Partners China, with participation from Redpoint Ventures, SenseTime Guoxiang Capital, Walden International, and L2F Lighthouse Founders’ Fund, according to Kr36. The company said it plans to use the capital to advance its GLOW generative world model, improve model generalization, and accelerate validation and production of its first product — an ultra-compact, fully foldable dual-arm household robot measuring about 40 centimeters called Knowin-X1.
According to the company, the Knowin-X1 is a roughly 50-kilogram, wheeled dual-arm robot designed for household use. It has 23 degrees of freedom, including two seven-degree-of-freedom arms, with a combined arm payload of 6 kilograms. The robot can fold to less than 40 centimeters in height for storage and uses cameras, depth sensing and Lidar for perception. Knowin is marketing it for tasks such as household organization, laundry and clothes folding, along with home monitoring and interaction.
The robot runs Knowin’s GLOW, or Generative Learning of World, embodied-AI system, which combines synthetic training-data generation, a 3D world model, vision-language reasoning, task planning and learned robot actions. The architecture uses cloud-based inference for its KnowinBrain foundation model while action commands are executed locally on the robot, with state and execution feedback exchanged at a stated 1 to 5 Hz, according to Knowin.
PokeBot
PokeBot closed a nine-figure Pre-A funding round, its second financing in three months since the embodied-intelligence startup was founded in April, according to Gasgoo. Shunwei Capital and Matrix Partners co-led the round, with participation from Ubiquant Ventures, Junshan Capital, SEE Fund, Liepin Investment, Yuannuo Capital and ETC Capital, while existing investors including Yunqi Partners, Xiaomi Strategic Investment, Highlight Capital, InnoAngel Fund and Oriental Fortune also participated. PokeBot previously raised tens of millions of dollars in an angel round led by Yunqi Partners.
The company is developing general-purpose robots for home and service settings, with a technology stack that combines a World Action Model, real-world reinforcement learning and proprietary data collection hardware. Rather than training separate models for individual tasks, PokeBot said its system is designed to learn physical cause-and-effect relationships and transfer capabilities across tasks, using real-world feedback to improve over time. The company has demonstrated long-horizon manipulation tasks including cooking Mapo Tofu, folding clothes, threading zip ties and tying sachets, according to Gasgoo.
Discover Robotics
Wuxi-based Discover Robotics closed a $100 million Angel+ round, coming less than a month after an initial angel financing that exceeded $100 million. The company, incubated out of Tsinghua University’s Intelligent Industry Research Institute and staffed with former DJI talent, focuses on consumer-grade embodied intelligence systems. It has developed hierarchical foundation models, data closed-loop systems, and high-fidelity simulators aimed at solving generalization and data scarcity challenges. The rapid successive raises have positioned it as a newly minted unicorn in the eyes of some market observers.
Discover Robotics sells the Airbot Play, a lightweight six-axis robotic arm built for embodied-AI research, mobile manipulation and data collection. According to the company, the arm weighs about 3.5 kilograms, requires no external control cabinet and runs on 24-volt DC power with a maximum draw of 240 watts. It can be integrated with wheeled robots, quadrupeds and other mobile platforms, and supports CAN- and RS485-connected end effectors as well as cameras mounted at the end of the arm. The software stack provides ROS/ROS2, Python, Web API and C++ interfaces, while supported simulation environments include Nvidia Isaac Sim, Gazebo and MuJoCo. Discover Robotics also says the platform can work with embodied-AI systems including ALOHA, Mobile ALOHA, RT-X and VoxPoser and can use large models to interpret natural-language instructions.
The company’s larger Airbot MMK2 Mobile Manipulation Kit 2 combines two six-degree-of-freedom lightweight arms with dexterous hands, multiple vision sensors, Lidar and a mobile base, along with nearly one meter of vertical lifting range. MMK2 uses a Sunrise 5 computing chip rated at 10 TOPS for sensor fusion, mapping, path planning, obstacle avoidance, object recognition and adaptive grasping. It also supports data collection through teaching devices, headsets and motion capture and is designed to run embodied-AI algorithms including ALOHA, OpenVLA and VoxPoser, making the platform both a robot research system and a way to collect and train the data needed for more general-purpose robot behavior.
Unitree
On Monday, Reuters reported that Unitree’s Shanghai initial public offering was oversubscribed about 8,289 times by retail investors, triggering a clawback mechanism that reallocated shares to the online tranche. The company said the final lot-winning rate for retail investors was about 0.018%, according to an exchange filing.
CNBC reported that Unitree priced its Shanghai IPO at 150.8 yuan a share ($22.34), valuing the humanoid and quadruped robot maker at about 61 billion yuan, or $9.04 billion, as it seeks to raise roughly 6.1 billion yuan on the STAR Market. The company plans to use the proceeds for robot software and hardware development, new products and manufacturing capacity, while its prospectus flagged U.S. restrictions as a risk to overseas growth after humanoid robots became its largest business in 2025.
Chinese Robotics Growth
According to data cited by both 36Kr and Global Times from the research platform ITjuzi, financing in China’s embodied intelligence sector reached 93.5 billion yuan, or $13.86 bullion, in the first half of 2026 across 322 deals, which is a five-fold increase from the same period a year earlier.
Global humanoid robot shipments reached about 19,100 units in the first half of 2026, up 272% from a year earlier, with Agibot overtaking Unitree for the top spot at 44% market share, according to Smart Analytics Global. Industrial and commercial applications accounted for more than 70% of shipments, while Chinese vendors supplied more than 97% of the global total and SAG forecast full-year shipments approaching 60,000 units and about $1.6 billion in industry revenue.
China’s 15th Five-Year Plan places robotics and embodied AI at the center of its industrial strategy, with the government aiming to shift AI research toward physical applications and expand commercialization of intelligent robots. The International Federation of Robotics noted in a May report that China already accounts for 54% of annual industrial robot installations worldwide, while the plan targets broader use of AI with traditional robots over the next five to 10 years and greater commercialization of humanoids toward the end of the 2026–2030 period.
