PsiBot raises more than $100 million for manipulation research and deployment

THE ESSENTIALS

PsiBot announced a round exceeding $100 million.
The company links a manipulation policy with an action-conditioned world model.
Industrial investors may provide development settings, but investment is not a customer deployment contract.

Industrial investors back manipulation research

PsiBot announced more than $100 million in new financing, with investors including Ningbo Tuopu Group, a Chery-backed fund, Lens Technology, 37 Interactive Entertainment, Wuhu Investment Holding Group and Fosun Fortune Capital. Existing investor Zhuhai Technology Industry Group also participated.

The company's own announcement identifies two spending priorities: further development of embodied world models and commercial work in logistics and advanced manufacturing. It emphasizes data collected from human actions as part of that strategy.

Two models have different jobs

PsiBot describes Psi-R2 as a policy that uses visual, language and robot-state inputs to plan actions. Its Psi-W0 model is intended to predict the consequences of candidate actions and help revise them. The company presents the combination as a way to improve manipulation in less structured environments.

The distinction is useful: choosing an action and predicting what that action will do are related tasks, but they are not identical. A robot attempting to pick up an object needs both an actionable movement and a way to judge whether the movement is likely to produce the intended result.

Factory access can help test the proposition

Several investors have manufacturing operations or supply-chain relationships. In principle, such relationships can make it easier to define useful tasks and gather examples from real work. They should not be counted as signed robot orders unless a deployment contract is separately disclosed.

For readers assessing progress, the relevant evidence would connect a named task to success rates, handling time and human intervention. The financing release does not supply a standardized cross-customer performance study. It establishes the resources and proposed technical direction; the commercial test is whether the system can repeat its work on objects and conditions it has not simply rehearsed.