Iván Hernández Dalas: X Square Robot secures $140M in funding for AI foundation models
The Quanta X2 semi-humanoid robot Source: X Square Robot
While artificial intelligence has advanced with large language models, or LLMs, robots need to understand the physical world to become general-purpose. X Square Robot yesterday said it has completed its Series A++ funding round, raising RMB 1 billion, about $140 million U.S.
“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,” stated Wang Qian, founder and CEO of X Square Robot. “This investment underscores shared confidence in our role as a catalyst for technological progress and will accelerate our expansion into high-value applications.”
Founded in 2023, X Square Robot said it is developing “general-purpose, end-to-end embodied AI foundation models.” The Shenzhen, China-based company, formally known as Variable Robotics Technology Co., added that its WALL-A system integrates vision-language-action (VLA) models with world models.
WALL-A supports continuous real-world evolution
“By using world models to predict actions and causal inference to understand feedback, the model significantly enhances the zero-shot generalization capabilities of robots performing mobile manipulation tasks in unstructured environments,” said X Square.
The company added that large-scale, real-robot reinforcement learning (RL) allows its foundation model to learn through physical interaction. It said this data-driven approach enables a robot to autonomously refine its skills and perform complex tasks in real-world environments.
Alongside the development of WALL‑A, X Square Robot introduced WALL‑OSS in September 2025. It said it designed the open‑source version of its model family to democratize embodied intelligence and accelerate community‑driven innovation.
“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,” noted Wang.
To win this race, X Square Robot said it is developing a closed-loop iteration of hardware, data, and models. The company claimed that it is the first one in China to scale up real-world data resources.
X Square has developed advanced data-capture tools, including teleoperation, exoskeletons, and the Universal Manipulation Interface (UMI). The company has also established a data pipeline to generate high-quality data at scale.
By using the foundation model to provide feedback on hardware design and data processing, X Square said it is improving acquisition efficiency and model performance, creating a self-reinforcing “flywheel.”
Quanta X1 delivers food using foundation models
X Square Robot said the adaptability of its systems proves that embodied AI has moved into practical, real-world deployment. The company recently demonstrated in autonomous food delivery in which its Quanta X1 wheeled bimanual robot, powered by WALL-A, successfully navigated indoor and outdoor tasks to complete a delivery in an open environment.
During the mission, Quanta X1 handled challenges such as strong winds, deformed packaging, and visual occlusions. Much like a human, the robot used the model’s causal inference to “fill in the blanks” when objects were partially hidden.
When encountering operational stalls or friction, the robot autonomously self-corrected and completed the task loop without any human intervention.
Quanta X1 also demonstrated this capacity in complex logistics. Facing piles of parcels, the robot used zero-shot generalization to identify irregular items. X Square noted that the model can unlock the potential of high DoF (degree-of-freedom) dexterous hands, enabling robots to master human-like skills ranging from tool use to precise card dealing. This conquers the “last centimeter” of precision manipulation, it said.
X Square goes from full-stack R&D to multi-scenario deployment
X Square said its hardware architecture was driven by model and data requirements. It has released two wheeled robots: the Quanta X1 and the Quanta X2 semi-humanoid robot.
In addition, the company said it has developed and used core components including robotic arms, joint modules, and controllers to be ready for mass production and commercial scale.
“X Square Robot continues to iterate across our three core pillars: models, data pipelines, and hardware,” said Wang. “By leveraging our technical depth and full-stack R&D, we consistently push the boundaries of robot performance.”
The company asserted that its systems are driving “a new era where embodied intelligence powers every layer of productivity.” It cited increasing deployments of its technologies across key industries such as advanced manufacturing, autonomous logistics, and senior healthcare services.
Strategic partners invest in X Square Robot
ByteDance and HongShan, participated in X Square Robot’s latest investment, along with several other strategic Chinese partners.
Previous rounds such as X Square’s $100 million Series A in September included backers such as Alibaba Group and Meituan. The company said it has demonstrated “compelling model capabilities and product potential that continue to attract investor confidence.”
“We’re honored to have the strong endorsement of our world-class strategic investors,” said Wang.
The post X Square Robot secures $140M in funding for AI foundation models appeared first on The Robot Report.
View Source
