What happened
TechNode reports that Chinese embodied-AI startup PsiBot completed a funding round of more than US$100 million. The reported participants include Ningbo Tuopu Group, a Chery Holdings-backed fund, Lens Technology, 37 Interactive Entertainment, Wuhu Investment Holding Group and Fosun Fortune Capital. Existing shareholder Zhuhai Technology Industry Group also continued investing.
TechNode reported on Aug. 28, 2026, that PsiBot, a Chinese startup focused on embodied AI and dexterous manipulation, had completed a new funding round worth more than US$100 million. The article identifies the financing as a completed round, but it does not state an exact dollar amount, the round’s formal stage, the company’s valuation or whether the amount includes debt, equity or other instruments. Those details remain unconfirmed by the source. In other words, the available account establishes the existence and broad purpose of the financing, while leaving the financial mechanics and the company’s resulting capitalization unspecified.
According to TechNode, the investors include Ningbo Tuopu Group; a fund backed by Chery Holdings; Lens Technology; 37 Interactive Entertainment; Wuhu Investment Holding Group; and Fosun Fortune Capital. TechNode also said that existing shareholder Zhuhai Technology Industry Group continued to invest. The report does not identify a lead investor, describe the ownership changes or disclose how much each participant contributed. Thus, the participant list cannot by itself show the round’s ownership impact or the relative importance of any investor.
TechNode said PsiBot plans to use the funding for research on embodied-world models, collection of data from human operations, and deployments in logistics and advanced manufacturing. The outlet also reported that the company is developing a dual-system architecture pairing its Psi-R2 operation-policy model with its Psi-W0 action-conditioned world model. The source provides no technical specifications, benchmark results, customer contracts, deployment schedule or independent confirmation of the system’s capabilities. The reported plans describe intended activities rather than completed outcomes. They also leave open how the research, data collection and deployments will be sequenced or connected.
Source details: technode.com ↗
Why it matters
The reported round links an embodied-AI developer with companies and investment groups connected to manufacturing, automotive and industrial activity. That combination could give PsiBot access to operational environments and data needed to develop systems that manipulate objects in the physical world, although the source does not independently establish any deployment results or commercial scale.
The financing is significant because it is tied directly to AI systems intended to act in physical environments rather than only generate text, images or code. Dexterous manipulation requires a system to connect perception, planning and control while coping with variations in objects, surroundings and timing. TechNode’s report indicates that PsiBot is investing in the research and data infrastructure needed for that work, but it does not show that the approach has solved those challenges. The physical setting makes the engineering problem especially sensitive to uncertainty: a model must respond to changing conditions while keeping actions controlled. No such results are reported here.
The participation of industrial companies and investment groups may matter more than the headline amount. Ningbo Tuopu Group and the Chery-linked fund give the round reported connections to automotive or manufacturing interests, while the stated deployment targets include logistics and advanced manufacturing. If those relationships lead to access to real operating environments, PsiBot could be positioned to test its systems on practical tasks. That is a potential consequence of the reported investor mix, not a confirmed outcome. The possible value of those connections therefore depends on access, scope and execution, none of which the article quantifies. The investor names alone do not establish that access.
The proposed use of human-operation data also raises questions about how physical-AI systems are trained and evaluated. Data gathered from people performing tasks can help capture motion and operating context, but the source does not explain what tasks will be recorded, how workers will be protected, whether the data will be representative, or how safety will be assessed. There is likewise no independent evidence in the article about productivity gains, displacement, error rates or the reliability of PsiBot’s models. These are open evaluation questions rather than findings from the reported round. The article supplies no answers about them.
What to watch next
Key unknowns include the round’s exact size, valuation, lead investor, financing terms and closing date. TechNode does not provide performance results, customer names, deployment locations, timelines or evidence that PsiBot’s models are already operating commercially. Further reporting should examine whether the planned data collection and industrial deployments produce measurable improvements in reliability, safety or labor productivity.
The first verification priority is the financing itself. Further reporting should establish the exact amount, financing structure, valuation, investor roles and date of completion through company filings, investor disclosures or direct statements. TechNode’s report is concrete about the participants and intended use of funds, but those financial details are not supplied and were not independently confirmed here. Without those disclosures, the headline amount cannot be compared precisely with other financings or used to infer the company’s financial position. The reported investor list does not resolve that gap.
The next question is whether PsiBot’s dual-system design moves beyond development claims. Psi-R2 and Psi-W0 are named in the report, but TechNode does not describe their architecture, training data, hardware requirements, task coverage or evaluation results. Useful follow-up evidence would include reproducible tests across varied objects and workspaces, failure rates, intervention rates, safety controls and comparisons with existing industrial automation. Those measures would help distinguish a research prototype from a system ready for repeated workplace use. The article does not provide that distinction.
Finally, observers should track the promised logistics and advanced-manufacturing deployments. The source does not name customers, facilities, countries, tasks or launch dates, so there is no basis to say that broad commercial availability exists. Reporting should determine whether the deployments are pilots or production systems, who remains responsible for decisions and physical safety, how human-operation data is governed, and whether the technology delivers measurable public or workplace benefits. The distinction matters because an announced intention to deploy is not evidence that a deployment has begun or succeeded. Those milestones remain to be documented.