The torrent of capital flowing into embodied intelligence is shifting decisively from robot hardware to the software "brain" that powers it. Qianjue Tech has announced the completion of a multi-hundred-million-yuan Series A+ funding round, with participation from Yuanhe Houwang, Xiaoguang Capital, Xineng Venture Capital, Inno Angel Fund, Jingming Capital, and Future Frontier Ventures.

This latest injection arrives just three months after the company closed its Series A round in June, underscoring the accelerating pace of investor enthusiasm for cognitive AI platforms in robotics.

Qianjue Tech is hardly alone in this trend. Extreme Vision, a company focused on world models, has completed multiple funding rounds within a single year, accumulating tens of billions of yuan in total financing. During the WAIC conference in July, its CEO Huang Guan revealed that its latest Pre-IPO round valued the company at $3 billion. Meanwhile, Manifold Space and Daxiao Robotics, both specializing in embodied intelligence models, are also aggressively raising capital.

The underlying driver is a new bottleneck gripping the robotics industry. Over the past few years, China's mature supply chain has dramatically reduced costs for motors, reducers, joints, and even complete robot bodies. Yet as robots enter homes, retail outlets, and service environments, the far more intractable challenge has emerged: enabling machines to understand dynamic real-world contexts, anticipate what happens next, and sustain decision-making and execution in the face of unpredictable changes.

Qianjue Tech's core technological approach centers on a predictive world model. Rather than directly generating actions from visual inputs, this framework enables robots to first comprehend the current environmental state, then simulate potential outcomes of different actions before determining the optimal next step.

For instance, when a robot needs to move a cup from a table to another location, simply identifying the cup is insufficient. The system must also predict how the object's state changes after grasping and, if a person moves the cup mid-task, recalibrate its plan based on the updated environment.

This technical route also tackles a crucial scalability question: how to make a single "brain" adapt across diverse robot platforms. Humanoid robots, wheeled robots, and quadruped machines each feature different sensors and actuation structures. If every new hardware configuration requires retraining a separate model, the economies of scale for model companies collapse.

Qianjue Tech is working to modularize, recombine, and upgrade the perception, decision-making, planning, and control capabilities of robots. The company has already advanced adaptation across humanoid forms, wheeled dual-arm systems, and quadruped robots, and is validating deployment in home, retail, and service scenarios.

As embodied intelligence funding pivots from "building bodies" to "building brains," the next phase will determine which players can genuinely help robots cross the usability threshold in the physical world.