Key takeaways
- Li Zexiang (李泽湘) is a Berkeley-trained roboticist and HKUST professor who mentored DJI founder Frank Wang and co-founded Googol Tech.
- In 2014 he launched XbotPark (松山湖国际机器人产业基地), a mentor-driven incubator that has spun out more than 80 robotics and hardware startups, at least six of them unicorns.
- His "build the founder, not just the product" model leans on the Greater Bay Area supply-chain (供应链) to compress prototype-to-mass-production timelines.
- The pipeline now feeds China's humanoid-robot (人形机器人) boom, training the engineers behind a new generation of bipedal machines.
- Honest gap: company counts vary by source, and the ecosystem's long-term hit rate is still unproven at national scale.
The lab that became a factory for founders
In a shared workshop in Dongguan, a white-shirted professor in his sixties hands a red-buttoned controller to a 22-year-old. The student is not writing a thesis. He is debugging a robot that could become a company within a year.
That professor is Li Zexiang (李泽湘), and the building is XbotPark (松山湖国际机器人产业基地), the robotics incubator he co-founded in 2014. The scene captures a bet that runs against Silicon Valley orthodoxy: that great hardware founders can be manufactured through a system, not just born with a spark.
From Berkeley to a Hong Kong classroom
Li was born in 1961 in Hunan province. He earned a PhD in electrical engineering and computer science at UC Berkeley in 1989, after undergraduate and master's work at Carnegie Mellon, and he later held posts at MIT's AI Lab and NYU's robotics lab.
He joined the Hong Kong University of Science and Technology (HKUST) in 1992, where he established the Automation Technology Center and, later, the Robotics Research Institute. In 1999 he co-founded Googol Tech (固高科技), one of China's earliest motion-control companies; it listed on Shenzhen's ChiNext board in August 2023.
The pivotal moment came when a graduate student, Frank Wang (汪滔), turned a helicopter-flight-control project into a company. That company became DJI, the world's dominant consumer-drone maker. Li became widely known as the mentor behind it.
Colleagues describe his teaching style as relentlessly project-based. Students are pushed to define a real product, then race it through supply chains instead of only writing papers. That pedagogy later became the DNA of XbotPark, and it is the reason so many of his advisees ended up running factories rather than staying in labs.
XbotPark: a system for making founders
With Berkeley colleague Ping K. Ko and Cheung Kong Graduate School of Business professor Jie Gan, Li launched XbotPark in 2014 in Songshan Lake, Dongguan. The model is deliberately anti-incubator. Instead of renting desks, it runs bootcamps, shares a prototyping factory, and drops founders into a five-minute walk of hundreds of component suppliers.
Official figures put the tally at more than 80 incubated robotics and smart-hardware companies, at least six of them unicorns. The portfolio includes robot-vacuum maker Yunji (云鲸), warehouse-robot firm Hairobot (海柔创新), and marine-motor brand ePropulsion. Li argues that the Greater Bay Area supply-chain (供应链) lets a team iterate far faster and far cheaper than a typical Silicon Valley path, because tooling, samples, and small-batch production sit inside one ecosystem.
Why this matters for humanoid robots
China's humanoid-robot (人形机器人) startups depend on exactly the dense supplier network and rapid-prototyping culture that XbotPark helped normalize. Unitree, Agibot, UBTECH and a wave of newer bipedal ventures all draw on Shenzhen–Hong Kong manufacturing density to move from render to running prototype in months.
Many of those companies' early engineers trained in the same Greater Bay Area ecosystem Li built, even when they were not directly incubated. His Shenzhen InnoX Academy and branch bases in Ningbo, Chongqing, Guangzhou and Hong Kong now replicate the formula, feeding talent into bipedal and embodied-AI (具身智能) ventures.
For global readers, the takeaway is practical: the cheap, fast hardware pipeline is a big reason Chinese humanoid prototypes reach market faster than many well-funded Western labs. It is less about one genius founder and more about a repeatable supply-chain machine.
A bipedal startup in this ecosystem can source motors, bearings, controllers and carbon-fiber shells within a day's drive, then revise a design overnight. That loop is one overseas competitors often take weeks to match, and it explains why so many humanoid (人形机器人) demos out of Shenzhen look production-adjacent rather than purely academic.
The education bet behind the hardware
Li has argued that the bottleneck is not capital or ideas but trained founders, so he has poured energy into what Chinese media call "new engineering" education. He founded the Shenzhen InnoX Academy and launched project-driven classes that treat a working prototype as the final exam.
The wager is that if you can systematize how a 22-year-old learns to ship hardware, you no longer depend on the occasional rare founder. Critics note the approach leans heavily on the Greater Bay Area's accidental supply-chain density, and that not every student becomes a Wang or a unicorn. Even so, the model has become a template other Chinese cities are now copying.
Honest limitations
- Company counts differ by source. XbotPark's own site says "more than 80" incubated firms, while some Chinese profiles cite 160-plus; we use the conservative official figure.
- "Incubated" spans a spectrum from hands-on mentoring to loose affiliation, so not every company is a Li-led venture.
- The model's success is tied to the Greater Bay Area's unique supply chain, which is hard to clone in other regions.
- We did not independently audit the financials or survival rates of the portfolio companies.
What readers can do now
- If you build hardware, study XbotPark's "prototype-in-one-building" playbook: supplier density beats office perks when speed matters.
- Investors should track which XbotPark alumni move into humanoid (人形机器人) and embodied-AI (具身智能) rather than only watching finished products.
- Engineers eyeing robotics should consider the Greater Bay Area apprenticeship route, not only a traditional PhD, as a path into real shipping hardware.