X Square Robot, a China-based developer of general-purpose embodied artificial intelligence, said January 12, 2026 that it has completed a Series A++ funding round of about $140 million, or roughly RMB 1 billion, as it accelerates development of what it calls a “robot brain” foundation model for the physical world and expands commercialization across logistics, manufacturing and healthcare use cases.
The round drew participation from ByteDance and HongShan, alongside other strategic Chinese partners, according to the company. X Square Robot said the new capital builds on prior backing from major technology and consumer internet players including Alibaba Group and Meituan, reinforcing investor interest in foundation-model approaches to robotics that aim to generalize across tasks and environments.
Founded in December 2023, X Square Robot focuses on end-to-end embodied AI foundation models designed to control robots in real-world settings where conditions are dynamic and imperfect. The company’s core platform is a vision-language-action (VLA) model family branded WALL-A, which it describes as integrating VLA capabilities with “world models” that simulate and predict how actions will play out in the physical environment.
X Square Robot said this architecture is intended to improve robots’ ability to perform mobile manipulation tasks in unstructured spaces by combining prediction with feedback-driven learning. The company highlighted the use of causal inference techniques to interpret outcomes and refine behavior, positioning the system as a path to stronger zero-shot generalization—handling tasks and scenarios that were not explicitly trained.
A key part of the strategy is large-scale learning through physical interaction. X Square Robot said it uses real-robot reinforcement learning to let its models acquire skills from trial-and-error in the world, complementing simulation and offline training. In the company’s view, the competitive moat in embodied intelligence will increasingly depend on closed-loop data pipelines that continually refresh model performance through repeated deployments.
To support that closed loop, X Square Robot said it has built data capture tooling and workflows to generate large volumes of high-quality manipulation and navigation data. The company cited teleoperation systems, exoskeleton-based capture methods, and a Universal Manipulation Interface (UMI) as mechanisms to scale real-world training signals, while also using model feedback to influence hardware design and data processing for efficiency gains.
In September 2025, the company introduced WALL-OSS, an open-source release within its model family that it said is meant to broaden access to embodied intelligence research and speed community experimentation. X Square Robot positioned the open-source effort as a complement to its commercial roadmap, which centers on deploying proprietary systems in high-value industry environments.
On the product side, X Square Robot said it has developed two robot platforms to pair with its models: Quanta X1, described as a wheeled bimanual robot, and Quanta X2, described as a wheeled humanoid robot. The company said it is also building core components in-house—including robotic arms, joint modules and controllers—with an eye toward mass production readiness and scalable deployments.
As a proof point for real-world operation, the company pointed to a recent demonstration of fully autonomous food delivery using WALL-A on the Quanta X1 platform. In that test, X Square Robot said the robot navigated both indoor and outdoor settings and completed an open-environment delivery mission while handling disruptions such as wind, packaging deformation and partial visual occlusions. The company said the system can self-correct when progress stalls, closing the loop without human intervention.
Beyond delivery, X Square Robot said its models are being applied to more complex logistics scenarios such as parcel handling, where irregular shapes and cluttered layouts can undermine rigid automation. The company also claimed advances in high-degree-of-freedom dexterous manipulation, describing skills that range from tool use to fine-grained precision tasks as the models evolve through continued data collection and training.
X Square Robot said it intends to use the new financing to push further on its three-part roadmap—models, data pipelines and hardware—while expanding into “high-value applications” and increasing deployments across industrial verticals. The company also provided a contact email for inquiries.
KEY QUOTES:
“We’re honored to have the strong endorsement of our world-class strategic investors. At X Square, we believe the key to enabling robots to truly master real-world tasks lies in the ‘robot brain’—a foundation model for the physical world that parallels virtual LLMs to shatter generalization bottlenecks. This investment underscores shared confidence in our role as a catalyst for technological progress and will accelerate our expansion into high-value applications.”
“The next phase of competition in embodied intelligence is essentially a battle of foundation models built on data closed-loops and their capacity for model evolution.”
“X Square Robot continues to iterate across our three core pillars: models, data pipelines, and hardware. By leveraging our technical depth and full-stack R&D, we consistently push the boundaries of robot performance.”
Wang Qian, Founder and CEO, X Square Robot
