Alibaba's "AI Nostalgic Investment": Investing Hundreds of Millions Exclusively, the Founding Teams of the "Four Little Dragons" Return to a New Track and Secure 1.2 Billion

On November 14th, Yiou reported that the embodied intelligence startup Dexmal announced the completion of hundreds of millions of yuan in Series A+ financing, with Alibaba as the sole investor.

Previously, Dexmal had completed two rounds of financing. In March 2025, it completed 200 million yuan in angel - round financing, with investors including well - known institutions such as Legend Capital, Jiukun Venture Capital, and Qiming Venture Partners. The Series A financing was led by NIO Capital, followed by Hongtai Fund, Lenovo Capital and Incubation Group, Xichuang Investment, and Zhengjing Fund. The old shareholders, Legend Capital, Qiming Venture Partners, and Jiukun Venture Capital, made additional investments, with Legend Capital making an over - subscribed follow - up investment.

As of now, the cumulative amount of Dexmal's three rounds of financing is nearly 1.2 billion yuan. The funds from this round will be mainly used for the R & D and implementation of software and hardware technologies for intelligent robots.

Dexmal was founded in March 2025 and was registered in Chongqing and Beijing successively. It focuses on the R & D and implementation of embodied intelligence software and hardware technologies and is committed to endowing robots with universality, reliability, dexterity, and intelligence through end - to - end large - model technology, helping them become an important part of everyone's productivity.

Dexmal's core founding team comes from Megvii Technology, led by Fan Haoqiang, Zhou Erjin, and Wang Tiancai. The team has both top - notch academic backgrounds and more than ten years of experience in the implementation of AI - native products, with relevant accumulations in large - model technology and robot scenarios.

Tang Wenbin, the founder and CEO of Dexmal, graduated from the Yao Class of Tsinghua University. He was the co - founder and CTO of Megvii. He is a gold medalist in the National Informatics Olympiad and the first - ever gold medalist of the "Yao Award". He is also the fifth Chinese contestant to achieve the TopCoder Target and once won the sixth place (first in Asia) in the ACM/ICPC International Collegiate Programming Contest World Finals.

Fan Haoqiang graduated from the Yao Class of Tsinghua University. He joined Megvii in 2010 (the 6th employee) and became the director of the algorithm department of the Megvii Research Institute at the age of 25, leading the team to win first place in many top - conference competitions.

Zhou Erjin graduated from the Department of Electronic Engineering of Tsinghua University. He won gold and silver medals in the IOI. He joined Megvii in 2015 (the 12th employee) and led the research on face recognition and the R & D of the algorithm production platform.

Wang Tiancai has been deeply involved in the fields of AI and computer vision, leading the construction of multiple core technology frameworks, publishing more than 30 top - conference papers with a cumulative citation of over 3300 times, and winning many important internal awards at Megvii.

Information from Tianyancha shows that the top three shareholders of Dexmal are Yuanli Juhe (Chongqing) Information Technology Co., Ltd., Lingji Qiyuan (Chongqing), and Lingji Tuoyuan (Chongqing). The legal representative of Yuanli Juhe is Yang Mu, the co - founder of Megvii, and Ant Group is the largest shareholder. The actual legal representatives of Lingji Qiyuan and Lingji Tuoyuan are Tang Wenbin, the co - founder and CEO of Dexmal (currently the legal representative of Megvii Technology).

Shareholder information of Dexmal (Chongqing) Intelligent Technology Co., Ltd.

Shareholder information of Yuanli Juhe (Chongqing) Information Technology Co., Ltd.

Megvii Technology was founded on October 8, 2011. It is one of the leading enterprises in the field of artificial intelligence in China and is known as one of the "Four AI Dragons" (Megvii, SenseTime, CloudWalk, and Yitu). The company was jointly founded by three genius teenagers from the Tsinghua University Computer Science Experimental Class (Yao Class), Yin Qi, Tang Wenbin, and Yang Mu, who serve as the CEO, CTO, and senior vice - president of the company respectively.

According to media reports, Megvii Technology has raised nearly 14 billion yuan in cumulative financing but has ended its two IPO attempts on the Hong Kong Stock Exchange and the STAR Market. Currently, Megvii is in a stage of business contraction and strategic transformation. In January 2025, the founders Yin Qi, Tang Wenbin, and Yang Mu pledged all their equity in Megvii Technology to the Chongqing Industrial Investment Mother Fund.

Judging from the shareholder background and the origin of the core team, Dexmal can be regarded as a re - entrepreneurship project of the early founding team of Megvii Technology. It is worth mentioning that Megvii Technology has formed a long - term association with Alibaba and Ant Group, from technical collaboration to capital linkage.

Alibaba's investment in Dexmal this time is quite meaningful. This is not just a simple "project investment" but can be regarded as a strategic move of "investing in people", "investing in the context", and "investing in the future". It clearly shows that the experience, technology, and resources of the first - generation Chinese AI entrepreneurs are being inherited and iterated in more promising new tracks (such as embodied intelligence) through the form of "re - entrepreneurship". Alibaba has accurately captured this trend and deeply bound itself to top - level AI talents and their new undertakings through a capital link.

In terms of products, Dexmal has built a complete embodied intelligence technology stack covering large - scale embodied intelligence models, multi - mode control systems, robot hardware platforms, and cloud - based collaborative platforms, achieving end - to - end intelligent control from large models to robot hardware.

Dexmal's self - developed end - to - end multi - modal embodied intelligence large model MMLA can achieve intelligent generalization across scenarios, tasks, and configurations by deeply integrating the underlying capabilities of multi - sensors, large language models (LLM), vision - language models (VLM), vision - language - action models (VLA), and world models.

It is worth mentioning that Dexmal has recently been continuously promoting the development of the embodied intelligence industry with open - source power. First, it open - sourced the VLA toolbox Dexbotic based on PyTorch, providing one - stop scientific research services for practitioners. At the same time, it launched the open - source robot hardware product DOS - W1 (Dexbotic Open Source - W1), significantly lowering the threshold for using robots and improving the convenience of maintenance and modification.

In addition, Dexmal jointly released the world's first large - scale real - machine evaluation platform for embodied intelligence, RoboChallenge, with Hugging Face. Currently, Dexbotic (software), DOS - W1 (hardware), and RoboChallenge (standards) have formed a deep synergy, injecting impetus into the development of the embodied intelligence robot industry from multiple dimensions.

Looking forward, Dexmal said that it will accelerate the collaborative innovation of algorithm - driven, hardware design, and scenario - closed - loop in the field of embodied intelligence and speed up the implementation of general artificial intelligence in the physical world.

This article is from the WeChat public account "Yiou Network" (ID: i - yiou), written by Lu Yongli and published by 36Kr with authorization.