Hangjiashuo Robotics
Within a week, Litchi Lab, Basal Intelligence, LivSyn Robotics, and AgiSense Robotics have successively announced their financing. None of the four companies builds complete machines: one handles data, one handles models, one handles architecture, and one handles haptics—all are in the "infrastructure" business.
01. Four Deals in One Week: Capital Bypasses Complete Machines
Litchi Lab secured an angel round plus an angel+ round, totaling hundreds of millions of CNY (according to media reports). Founded by Li Yiming, an assistant professor at Tsinghua University, the company positions itself in Physical AI Infra—providing a complete set of infrastructure for data collection, governance, training, and evaluation for embodied models. Ant Group increased its investment in two consecutive rounds and led the angel+ round, with CMC Capital, Guoqi Investment, and funds under CICC following suit. Its products are the T1 data collection glove and the Ego data collection device, doing one thing: transferring human actions of manipulating the real world into datasets that models can digest.
Basal Intelligence secured a seed round just three months after its establishment, led by 5Y Capital, with participation from Ivy Capital, Linear Capital, and Huashan Capital. Founded in July 2026, the company was co-founded by eight researchers with Tsinghua and Peking University backgrounds, focusing on action-native embodied foundation models—treating "action" as the primary representation of the model, rather than an output layer added after training. Its cross-embodiment model X-VLA, using less than 1,000 hours of pre-training data, won the championship in the Agibot Challenge at IROS 2025 among over 400 teams.
LivSyn Robotics completed a Series A round of over 100 million CNY, with participation from Huafang Capital, Daohe Yuanqi, Suwen Electric Power, and others. Founder Yang Hongbing formerly served as the CTO of a leading humanoid robot company, and CTO Jiang Yuhua is a post-2000 Ph.D. graduate from Tsinghua University. The company is betting on the RUDA cross-embodiment unified architecture, enabling a single set of infrastructure software to adapt to different robot embodiments.
AgiSense Robotics secured a tens-of-millions-of-CNY angel round from Lihe Capital and Hefei Innovation Investment. This Shenzhen-based company develops dense force field visuo-tactile sensors—extracting force and geometry from the same imaging process. The Lux series has been delivered in batches, with the next step extending to electronic skin, 6D force sensors, and data collection systems.
02. Why Invest in "Infrastructure" Instead of Complete Machines
The answer lies in the valuations.
California-based FieldAI does not build a single robot but only sells unified robot "brain" software; market rumors suggest its valuation in the new round of financing is about USD 10 billion. Domestic X Square Robot closed four rounds within two months, with a post-investment valuation exceeding CNY 20 billion.
Looking at complete machines again: they compete on supply chains and delivery, with gross margins squeezed from both ends by costs and prices. A single robot is sold once, while a universal infrastructure can be installed in all robots—the valuation multiplier for a platform business is not in the same coordinate system as hardware.
Infrastructure has an even harder pivot: data. A frequently cited set of figures shows that the current global volume of high-quality physical interaction data is about 500,000 hours, while the demand for training general-purpose embodied models is measured in billions of hours, resulting in a gap of over 99%. Whoever controls the data pipeline stands upstream of model developers.
03. The Density of the Tsinghua Network: Coincidence or Signal?
Putting the lists of the four companies together: Li Yiming, an assistant professor at Tsinghua University; the eight-person team of Basal Intelligence comes from Tsinghua, with founder Li Jianxiong being the first Ph.D. student of Tsinghua AIR; Yang Hongbing and Jiang Yuhua also graduated from Tsinghua. Academic research achievements are bypassing the long chain of commercialization and directly turning into companies.
Embodied AI is one of the few tracks where the distance from "paper to product" is short enough. Professors leading students into the field is much faster than in other industries; for investors, a team that is both knowledgeable and young is an asset in itself.
04. Three Types of Investors, Three Different Calculations
Ant Group increased its investment in two consecutive rounds, targeting the entry point to the physical world—once robots can pick up objects and manipulate them, the interface for payment and offline scenarios will have a new owner.
5Y Capital led the investment in Basal Intelligence, betting on the technological route. Action-native and cross-embodiment architectures have not yet converged; entering the market at this time means betting on the position of a standard-setter.
State-owned and industrial players like Guoqi Investment, Hefei Innovation Investment, and Suwen Electric Power seek industrial chain positioning and interfaces with their own businesses—automotive, electric power, and local industrial funds all want to secure a seat at the infrastructure layer.
Competition in complete machines is intensifying, with profits thinning day by day; capital is placing its bets on the layer that "makes all robots smarter." The primary market for embodied AI has entered an "infrastructure arms race"—data and universality are becoming the core narrative of this round of valuations.
The next batch of high-valuation companies may not be the one that builds the most human-like robots, but rather the one that makes all robots smarter.
This article is compiled based on publicly available online information and is for reference only, not constituting investment advice.