One billion! Robotics' biggest financing of the opening year is born! Already gathered Ali Byte and other giants

mechanical personOutlook January 12, 2011 - Today, Shenzhen-based robotics creatorX Square Robotannounced that it has recently completedOne billion dollars.A++ roundfinancing.

This round of financing was provided bybyte-jumpingLtd, Sequoia China, Beijing Information Industry Development Fund, SZVIC, Nanshan Battle New Investment, Tin Venture Capital and other top investment institutions and diversified local platforms jointly invested in the company. It is also reported that this is the first investment of SZVIC AI Fund since its establishment.

Founded in December 2023, Self-variable Robotics focuses on self-researching generalized embodied intelligence big models, using real-world data as the main data source to build generalized robots with fine operation capabilities, and according to its official website, it is one of the first companies in China to adopt a fully end-to-end path to realize a generalized embodied intelligence big model.

Previously, the company has picked up several large rounds of financing, and last September, it closed around nearlyOne billion dollars.Series A+ financing byAlibaba CloudThe investment was led by the company's investment arm, CST Capital.Meituan's war chest, Lenovo Star and other followers; in May last year, it closed theseveral hundred million dollarsSeries A financing, led by Meituan Battle Capital and followed by Meituan Dragon Pearl.

The company has consecutively takenByteDance, Meituan, AliThree major Internet companies have injected capital and are also currentlyThe only one in China to receive investment from these three major manufacturersof embodied intelligent enterprises.

Qian Wang, Founder and CEO of Variable Robotics, graduated from Tsinghua University and was one of the first scholars in the world to introduce attention mechanisms in neural networks. During his PhD, Qian Wang was involved in several Robotics Learning studies at top robotics labs in the US, with his research covering a wide range of cutting-edge areas in robotics.

联合创始人兼CTO王昊是北大计算物理博士，在粤港澳大湾区数字economics研究院（IDEA研究院）期间担任封神榜大模型团队算法负责人，发布了国内首个多模态开源大模型“太乙”，首批百亿级大语言模型“燃灯”以及千亿级大语言模型“姜子牙”。

▲Wang Qian, Founder and CEO of Autovariable Robotics

At the hardware level, Autovariable Robotics Design released the Universal Wheeled Dual Arm RobotQuantum OneA fully self-developed wheeled two-armed humanoid robot adapted for multimodal large model controlQuantum IITwo high-performance robot bodies.

Among them, Quantum 2 has 20 degrees of freedom in one hand, which can sense subtle pressure changes. At the same time, based on the arm and hand integrated exoskeleton technology, the first self-transformation “humanoid mechanical arm + high degree of freedom dexterous hand” integrated whole-body telemanipulation program, so that Quantum 2 not only can collect high-quality data to feed the model training, but also will be deeply integrated with the self-research model, and really enter the real world scene of landing applications.

▲Quantum II (Source: Self-variable Robotics official microblog)

At the modeling level, autovariable robotics has developed its ownWALL-A modelThe core architecture of WALL-A is the first system paradigm that deeply integrates VLA with the world model. As a native multimodal input-output architecture, WALL-A is the first to realize embodied multimodal thought chain.WALL-A utilizes the world model mechanism for spatio-temporal state prediction, collaborates with visual causal reasoning to understand the environmental feedback, and internalizes the physical common sense from the data through the learnable memory mechanism.

▲Based on the WALL-A model, the robot “Xiaoliang” learns to independently make scented sachets in just a few days (Source: official microblog of Variable Robotics)

In addition, the company adheres to the closed-loop iteration of hardware-data-model, and it has self-researched a variety of data mining equipment such as master-slave teleoperation, exoskeleton, and ontology-less, and has built a model-driven data pipeline, which continues to generate scaled and high-quality data through data generation, data filtering, data enhancement, and data labeling.

Now, on the basis of adhering to the full-stack self-research of hardware and software, Variable Robotics has defined the hardware architecture of the robot in depth from the model algorithm and data-driven demand, and designed and released two high-performance robots, “Quantum One” and “Quantum Two”. On the basis of modeling and data-driven needs, we have defined the hardware architecture in depth, designed and released two high-performance robots, "Quantum I" and "Quantum II", and realized the comprehensive self-research and in-depth algorithmic adaptation of core components such as robotic arms, joint modules, power drivers, and main controllers, which has contributed to the significant reduction of the cost of the whole machine.

Its products have gradually entered a number of high-value fields such as industrial manufacturing, logistics, and senior care.