DexForce Uses Virtual Data to Train Humanoid Robots, Raises $147M at $1.4B Valuation

DexForce Tackles Humanoid Robot Training Challenge

Chinese startup DexForce is addressing one of the humanoid robot industry's most pressing obstacles by training humanoid robots using virtual data instead of relying on expensive real-world data collection1

1

. The Shenzhen-based company employs synthetic data generated in virtual environments to teach machines how to perform real-world tasks without human assistance, offering a cost-effective method to train humanoid robots2

2

. At DexForce's offices, screens display virtual robots repeatedly placing objects on tables while variables like arm angles, table colors, and object types continuously change3

3

.

Why Traditional Training Methods Fall Short

Chief Executive Jia Kui, a machine-learning professor at the Chinese University of Hong Kong, Shenzhen, explained that achieving physical artificial general intelligence through purely real-world data collection would require 10 trillion samples1

1

. He told Reuters that manual data collection for infinite human tasks is fundamentally inefficient, especially when human trainers may need to repeat a movement more than 100 times to generate precise examples2

2

. While China has developed humanoids capable of outperforming humans in running and jumping, the harder commercial test lies in whether they can reliably carry out useful work in factories, offices, and homes1

1

.

The AI-Driven Training Methodology Behind DexForce

DexForce develops AI software designed to simulate physical environments and models intended to predict how objects move and interact3

3

. The company creates a digital version of a workplace or task, trains a robot in the simulation, and then transfers that learning to physical machines1

1

. This AI-driven training methodology represents a stark contrast to the dozens of humanoid robot data training centers currently being built nationwide at great expense, which are often heavily reliant on government subsidies2

2

. Most Chinese robot makers use a combination of simulated and real-world data, with many citing access to high-quality training data as a key bottleneck1

1

.

Strong Backing and Commercial Deployment

Founded in 2021 by Jia Kui, the roughly 500-employee startup is backed by Lenovo Capital, the investment arm of Lenovo, and Lens Technology, as well as several state-owned funds2

2

. DexForce raised 1 billion yuan ($147 million) in a recent Series B funding round at a valuation of more than 10 billion yuan3

3

. Bloomberg reported the company was considering a Hong Kong IPO, though Jia declined to comment1

1

. The firm's software is already deployed in industrial settings by major manufacturers including Toyota, Samsung, Panasonic, Foxconn, Haier, and Hisense2

2

. Beyond factories, DexForce's humanoid robots make coffee and serve customers at malls and tourist sites in more than 15 cities nationwide3

3

.

Implications for the Robotics Industry

Jia Kui believes this approach could more quickly power a breakthrough in robot capability, akin to the one ChatGPT triggered after its 2022 release helped propel generative AI into the mainstream1

1

. Reaching physical artificial general intelligence will require AI systems that can perceive and navigate their surroundings and adapt to variation—a technological milestone where a humanoid can complete new tasks in an unfamiliar environment on command2

2

. Watch for how DexForce's synthetic data approach influences the broader robotics industry's training methodologies and whether it can deliver on the promise of achieving true task versatility at scale.