"Variable Robotics" Closes Two Consecutive Rounds Worth Hundreds of Millions of Yuan, Yunqi Capital Participates | Yunqi Capital

Today (May 12), embodied intelligence company "X Square Robot" announced it has completed a Series A financing round of several hundred million RMB, with exclusive investment from Meituan. Earlier, the company closed a Pre-A+++ round at a similar scale, with participation from Yunqi Capital, Meridian Capital, and GF Xinde Investment Management. To date, X Square Robot has completed seven financing rounds totaling over one billion RMB.

Founded in 2023, X Square Robot is described as the earliest domestic company to pursue a fully end-to-end path toward a general-purpose embodied intelligence foundation model. Since late 2024, the company has also made breakthroughs in multimodal output and embodied chain-of-thought capabilities.

This edition of "Yunqi Capital" brings you details on X Square's latest round and its technical and product progress.

Chen Yu, Partner at Yunqi Capital

X Square is the team we've encountered in the embodied intelligence space with the highest "concentration" of foundation model expertise. Since its founding in 2023, the team has remained steadfast in exploring the scaling law of embodied intelligence and building an end-to-end manipulation foundation model. Today, this technical approach has been widely recognized by academia and industry as one of the most promising directions for breakthroughs. Yunqi began tracking X Square immediately after its founding and made our investment decision after seeing the team's generalization demos — and more importantly, the deep capabilities behind them in large models, data engineering pipelines, and foundational infrastructure. The core driver of embodied intelligence is AI, and we look forward to X Square's continued breakthroughs in general foundation models, achieving the embodied ChatGPT moment as soon as possible.

The following is excerpted from 36Kr:

This Series A round was led by Meituan's strategic investment arm, with Meituan Longzhu participating as a co-investor. The proceeds will be used to continuously accelerate the synchronized iteration of X Square's fully self-developed end-to-end general-purpose embodied intelligence foundation model and robot hardware, as well as to develop and deploy intelligent solutions across multiple application scenarios.

X Square Robot (X Square Robot) was founded in December 2023, focusing on the development of a "general-purpose embodied foundation model." Using real-world data as its primary source, the company aims to build general-purpose robots with fine manipulation capabilities.

The core team brings together experts from world-renowned AI and robotics labs and scholars from top universities at home and abroad, including researchers who were among the earliest globally to propose the Attention mechanism (the core of the Transformer architecture), experts from top international robotics labs, technical leads behind China's first trillion-parameter large models, and seasoned veterans in robot hardware.

As the earliest domestic company to adopt a fully end-to-end path for general-purpose embodied intelligence, X Square Robot is committed to building a unified architecture integrating perception, planning, and control. The company's self-developed "WALL-A" model, part of the "Great Wall" manipulation foundation model series, features multimodal information fusion capabilities, integrating diverse types of perceptual data, natural language instructions, and motion control signals to achieve end-to-end input-to-output mapping.

Based on large-scale general knowledge pre-training and multi-task learning mechanisms, the "WALL-A" model has already demonstrated zero-shot generalization capabilities in some previously unseen task scenarios — broad zero-shot generalization being one of the key hallmarks of achieving general-purpose robots.

Meanwhile, X Square Robot has built a model-driven data closed-loop system. The company has independently developed dozens of data processing models and multiple generations of data collection equipment to enable automated quality control and substantially improve data collection efficiency.

On the hardware side, the company has independently developed and continuously optimized robot hardware specifically designed for multimodal foundation model control, better meeting the demands for fine manipulation and stable operation in open environments. Currently, X Square's robot hardware has been deployed in multi-step complex task scenarios.

Notably, since late last year, X Square Robot has made breakthroughs in multimodal output and embodied chain-of-thought — ahead of Gemini Robotics' chain-of-thought research announcement in March of this year. The company's model architecture is centered on multimodal output, achieving end-to-end information fusion, enabling efficient alignment between visual, linguistic, and action information channels, significantly enhancing the model's contextual reasoning and self-feedback capabilities in ultra-long sequence tasks. The company has also continued to make progress in dynamic environment perception, real-time task planning, and ultra-long-horizon tasks. These breakthroughs and advances have further improved robots' autonomous decision-making and execution efficiency in complex open environments.

Wang Qian, Founder and CEO of X Square Robot, believes that: in the near term, embodied intelligence development depends on model algorithm advantages; in the medium term, on data advantages; and in the long term, the core lies in product advantages. X Square has consistently deepened its accumulation across these three directions, and has taken the lead in deploying and practicing in open service scenarios domestically, exploring service closed-loops in different contexts.