Li Qiang, CTO of Cainiao, has started his own business and received investment from the angel investors of Minimax.

Today, ChinaVenture exclusively learned that Li Qiang, former CTO of Cainiao, has officially embarked on his entrepreneurial path and founded Hangzhou Quantum Dynamics Intelligence Co., Ltd. So far, the company has closed a seed round financing of over 100 million RMB, with investors including leading funds such as Yunqi Capital, Guoxiang Capital, as well as top-tier industrial capital.

Prior to this, the logistics industry where Li Qiang used to work was facing an awkward gap: on one hand, there was the harsh reality of difficult recruitment and continuously rising labor costs, on the other hand, the previous generation of automation technology could never cross the threshold when dealing with hand-eye coordination and flexible operation scenarios. As a result, this team spun off from Alibaba Group came into public view.

It is worth mentioning that the United Nations once reported on Li Qiang's technology in previous years, and this entrepreneurship is quite equivalent to a new restart for Li Qiang.

Recently, ChinaVenture had an exchange with this newly founded entrepreneur. During the conversation, Li Qiang stated that Quantum Dynamics is positioned as a Physical AI system company, which cuts into the most challenging "hand-eye coordination" operation in logistics scenarios. By self-developing an embodiment-native world large model, the company aims directly at the fundamental pain point of low efficiency of "replacing human workers with machines" in sorting scenarios, and strives to build a robot fleet that can be deployed on a large scale by solving the problem of flexible operation in unstructured environments.

"Whoever takes the lead in closing the data flywheel loop in the real physical world will hold the ticket to the Physical AI era," Li Qiang said during the exchange. This means that the window for PMF (Product-Market Fit) verification in logistics scenarios has quietly arrived.

Former Cainiao CTO Launches Startup, Targeting the Untapped "Hand-Eye Coordination" Vacuum Zone in Logistics

In early 2026, Li Qiang, who had more than 10 years of in-depth experience at Cainiao, made a decision that surprised the whole industry: to start his own business.

From joining Cainiao in 2015 to starting to explore warehouse automation in 2016, Li Qiang said he has personally witnessed the whole process of the previous generation of technology growing from 0 to 1. But after nearly 10 years of practice, this CTO found a harsh fact: the old technology can only solve the storage and handling of standard containers and standard boxes, and almost all operations involving hand-eye coordination — parcel weighing and waybill pasting, parcel flipping and feeding, restocking and shelving, picking and off-shelf, cartoning and packaging — can hardly be handled properly.

"With the previous generation of technology, we can only solve 20% to 30% of the problems, a large number of problems remain unsolved," Li Qiang said. It was not until 2025 that he clearly realized that the problems that had been unsolvable for the past 10 years finally found a solution with the new technology today.

This "new technology" is Physical AI.

Few people can match Li Qiang's understanding of logistics scenarios. During his tenure in charge of Cainiao's technology, the electronic waybill system he led was once the infrastructure of China's logistics — once it malfunctioned, the entire express delivery and logistics industry across China would face major disruptions. This awe of "large-scale deployment" and deep understanding of the stability bottom line of large systems became the original intention of his entrepreneurship.

Up to now, centered on Li Qiang, Quantum Dynamics has assembled a top-tier team spanning technology, product and industry sectors. Among them, founder Li Qiang has previously served as CTO of Cainiao Group and CTO of Alibaba's International Digital Commerce Division, managing thousands of R&D team members distributed around the world; he once concurrently served as General Manager of Cainiao Autonomous Driving Vehicle, taking charge of both technology R&D and business expansion, and was the top leader responsible for both the implementation and commercialization of autonomous driving technology.

The CTO of Quantum Dynamics used to be the head of online algorithm at a leading autonomous driving company in the industry, with in-depth experience in the full-stack technology of autonomous driving; the chief scientist is a PhD from a top domestic university, with visiting scholar and postdoctoral experience at well-known North American universities and renowned embodied intelligence laboratories.

In addition, the technical leader of the humanoid robot division of a listed company serves as the hardware head, who has focused on the field of robotic arms and dexterous hands throughout his undergraduate, master and doctoral studies. Besides, the head of commercialization graduated from the Department of Physics, University of Maryland, and has worked as an algorithm expert at Goldman Sachs, Amazon and Meta, with 10 years of entrepreneurial experience in logistics scheduling systems, and has matured customer and partner resources at home and abroad including Lenovo China and Lotte Korea.

The in-depth integration of this cross-background talent team allows Quantum Dynamics to break away from the limitation of only making breakthroughs in a single technical point from the very beginning. In their view, the implementation of Physical AI is not a breakthrough of a single technical point, but a competition of systematic capabilities — from the upper-layer scheduling system to model training, from data flywheel to end-side deployment, none of the links is dispensable.

Embodiment-Native World Model + Data Flywheel, Seizing the "Swarm Intelligence" Track of Physical AI

Then, specifically in the logistics scenarios, what exactly will Quantum Dynamics do to solve the most difficult problems? In short, it is a full-stack solution covering the robot body, the embodiment-native world model, and the fleet scheduling system.

Generally speaking, most players in the industry that develop VLA models only have a "brain" but no "hands and feet", while most teams working on embodied intelligence only demonstrate single-point capabilities. The difference of Quantum Dynamics is that it aims at the "system" from the very start.

The first is the technical route of the world model. During the conversation, Li Qiang said that the team will start from scratch, completely rebuild the system, and directly target the embodiment-native world model, rather than making patchwork modifications on the video generation model. This model integrates multiple modalities such as vision, tactile sense and force feedback, and has three capabilities at the same time: WorldModel (predicting the evolution of the world), World Action Model (outputting actions) and World Value Model (evaluating state value).

The second is the first-mover advantage of the data flywheel. Back when Li Qiang's team was at Cainiao, they operated China's first L4 autonomous driving fleet with a scale of thousands of vehicles, and established a complete system for automated data collection, labeling, training and evaluation. In the Physical AI era, the real machine interaction data generated by robots deployed on a large scale in the real physical world is undoubtedly the data asset with the highest value and the thickest barrier.

"In the Physical AI era, data will contribute more than 50% of the value, even higher than in the large language model era," Li Qiang judged. But high-quality real machine operation data is never a public resource that can be bought with money — only by deploying robots to the real physical world, letting them work, make mistakes and get corrected on the front line, can high-value data that conforms to the real distribution be generated.

The last part is the upper-layer fleet scheduling system. In Li Qiang's plan, what Quantum Dynamics delivers is never an isolated robot, but an organic robot fleet — how to assign tasks, how robots cooperate with each other, and how to let the whole fleet share capabilities after one robot learns new skills, all of which are included in the scope of this system.

The core logic of this full-stack solution lies in: the real bottleneck of large-scale deployment of Physical AI is not single-machine intelligence, but swarm intelligence. In Li Qiang's words, it is not rare that one robot can complete picking tasks in a warehouse, what is rare is that hundreds or thousands of robots work in the same warehouse at the same time.

From Domestic to Overseas, How Far Is the "GPT Moment" of Physical AI?

Of course, as a CTO who started a business after leaving a large tech company, Li Qiang not only has theoretical technical reserves, but Quantum Dynamics also has a clear commercialization strategy. This is probably an important reason why Quantum Dynamics has received support from Yunqi Capital, the angel investor of Minimax, and SenseTime, one of the "Four AI Dragons" in China, at the seed round stage.

In terms of scenario selection, Quantum Dynamics follows an evolution path from "closed" to "open". That is, it starts from the warehouses and distribution centers of B2C logistics, then moves to semi-open outlets and stores, and finally enters the fully open end distribution scenarios. It is worth mentioning that in this process, Quantum Dynamics does not pursue a "one-size-fits-all" universal solution with a general-purpose robot that can handle all scenarios. Instead, it uses robots of different configurations on the same platform for different scenarios, because letting the right robot do the right thing is the best strategy.

In terms of customer expansion, Quantum Dynamics' layout covers both domestic and overseas markets. According to Li Qiang, the company has already contacted a number of leading logistics enterprises, including a leading domestic comprehensive logistics enterprise, and Quantum Dynamics is about to sign a strategic cooperation with them, covering picking pilots in multiple core warehouses across the country. At the same time, the world's largest health care product e-commerce platform has also confirmed that it will adopt Quantum Dynamics' picking solution in its overseas warehouses, and the first batch of projects is planned to be deployed within this year.

As for the expansion rhythm of the overseas market, Quantum Dynamics has its own pace — prioritizing Japan and South Korea, followed by Europe and the United States. The core reason for this priority is that the urgency of labor shortage determines the strength of market demand; in addition, Japan and South Korea are relatively close to China; furthermore, the controllability of data compliance also determines the speed of implementation.

From domestic to overseas, from ecological partners to external customers, Quantum Dynamics has taken a more pragmatic path in the implementation of commercialization — that is, from the very beginning, it is oriented to real demands and real scenarios.

Looking ahead, Li Qiang said that by the end of 2026, the team hopes to reach the PMF level in at least 1 to 2 scenarios; in 2027, the goal is to let more than 1,000 robots work in the real physical world every day.

Looking back at the Internet era, massive text data gave birth to ChatGPT. Now entering the Physical AI era, humans need massive physical data that conforms to the real physical laws, so that robots and intelligent manufacturing can be truly implemented. In this regard, Li Qiang believes that the GPT moment of generative Physical AI may come earlier than most people expected.

What Quantum Dynamics wants to do is to turn every grasp in non-standard scenarios, every failure, and every recovery in the most realistic training ground of logistics into data nutrients that drive AI evolution, and become the one that brings the GPT moment of embodied intelligence one step closer.

This article is from the WeChat official account "ChinaVenture", author: Chen Mei, authorized for release by 36Kr.