Shenzhen-based DexForce is using virtual simulations to train humanoid robots, reducing the need for costly manual data collection. Valued at over 10 billion yuan, the firm counts Toyota and Samsung among its clients. As it weighs a potential Hong Kong IPO, its ability to move from virtual training to real-world factory performance will be the key test for investors.
Shenzhen-based robotics startup DexForce is pushing for a new way to train humanoid robots by moving away from traditional, manual data gathering. Instead of recording humans performing tasks one by one—a slow and expensive process—the company uses synthetic data generated in virtual environments. This digital approach allows robots to learn from countless scenarios, such as changing light, different object shapes, or varying task angles, in a fraction of the time required by physical data collection.
This shift in methodology addresses a major bottleneck in the robotics industry: the ability of machines to adapt to unpredictable real-world environments. By using software to simulate these conditions, DexForce aims to accelerate the development of physical artificial intelligence. The goal is to create robots that can eventually handle assembly line tasks with the same flexibility as humans. This tech-first strategy has attracted interest from large industrial players, with the company’s software already in use by major manufacturers including Toyota, Samsung, Panasonic, Foxconn, Haier, and Hisense.
From a financial perspective, the company has scaled rapidly since its founding in 2021. Following a recent Series B funding round, its valuation has crossed the 10 billion yuan mark. The firm is backed by significant institutional investors, including Lenovo Capital and Lens Technology, which signals strong backing for its role in the industrial robotics supply chain. Reports suggest that the company is now considering an initial public offering in Hong Kong, marking a significant milestone in its growth journey.
However, investors may want to monitor the execution risk associated with this business model. While synthetic training is faster and cheaper, the main challenge for any robotics startup is the 'reality gap'—the risk that software trained in a virtual world may behave differently when placed on a physical factory floor. The true test for DexForce will be its ability to consistently demonstrate that its virtual training translates into reliable, safe, and efficient performance in actual factory settings. As the company prepares for a potential public listing, the ability to scale this technology without compromising on accuracy will be a critical factor for long-term viability. Investors may keep an eye on project timelines and client retention as the company attempts to expand its footprint in global industrial markets.