The reported financing for DexForce raises one practical question about its robot-training claim: can behavior learned in a simulated workplace or task transfer to physical machines? It is the question that gives the round its real significance. Reuters reported the financing alongside DexForce’s approach; evidence from physical workplaces is still needed to assess performance.
Reuters reports that Shenzhen-based DexForce raised a 1 billion yuan Series B at a valuation above 10 billion yuan. Reuters also reported the company’s stated sequence: create a digital version of a workplace or task, train a robot in simulation, and transfer the behavior to physical machines. That reported sequence is the proposition attached to the round, with independent workplace evidence still needed to assess the transfer.
DexForce’s own official core product page describes DexVerse, DexWorldModel, and EmbodiChain. Its W1 Pro page describes a Sim2Real VLA pipeline. Those pages identify products and a pipeline that DexForce says it offers. Reuters’s report, meanwhile, records the financing and the company’s account of its training approach. The distinction is important because an official product description and independent workplace evidence are different kinds of information.
DexForce’s Simulation-to-Physical-Machine Claim
Reuters reported DexForce’s claim that simulated data can be faster and cheaper than repeatedly collecting demonstrations from human trainers. In the company’s framing, preparation occurs in simulation before a behavior is moved to a physical machine. That is a clearly defined comparison, but it remains DexForce’s claim, reported by Reuters. Independent evidence is still needed to assess the claimed speed or cost result.
The reader-facing consequence is simple. DexForce has framed a route from simulation to physical machines; outside measurement in a workplace remains the relevant test of the physical result. This is not a broad verdict on the company’s products. It is the direct test of the handoff described in Reuters’s report.
The question will be familiar to readers of RobotEra’s VPP2 Wins a Simulation Lead, Then Faces the Harder Test. DexForce has its own reported approach and its own financing, while physical-workplace evidence remains the relevant standard for judging the transfer it describes.
Reuters also reported that founder and chief executive Jia Kui has argued that the scale of the training problem can reach 10 trillion samples. Jia presents that as an estimate rather than an industry measurement. It explains why DexForce presents simulated data as an alternative to repeatedly gathering demonstrations from human trainers. Independent workplace evidence is needed to establish any training requirement, saving, or result.
What Reuters Reported About DexForce’s Scale
Reuters identified Jia as DexForce’s founder and chief executive and reported that the company has roughly 500 employees. Those details make the responsible actor clear: DexForce, led by Jia, is the company advancing the reported simulation-to-physical-machine approach. Reuters’s reported valuation above 10 billion yuan belongs to the same financing account. It is a reported valuation and is separate from evidence on reliability, cost, safety, or transfer in a physical workplace.
Reuters also reported company and customer-use information, including deployment claims. Those claims are company information reported by Reuters, rather than independently audited evidence of adoption. That qualification belongs at the center of the story because a reported deployment claim and an independently observed physical-workplace result answer different questions. The first tells readers what Reuters reported from company and customer information; the second would test the reported training approach under the conditions where it is meant to operate.
The same Reuters account brings funding, valuation, workforce size, and reported customer-use information together with DexForce’s training claim. They identify the company and the proposition behind the round, while leaving the physical-workplace test where it belongs: in independently observed results.
The company pages remain useful within their limits. DexForce describes DexVerse, DexWorldModel, EmbodiChain, and a Sim2Real VLA pipeline. Those labels identify the company’s stated product framework, while independent workplace evidence is still needed on reliability, cost, safety, and simulation-to-real transfer.
Readers looking across China’s robotics sector can also follow Blue Worm Launches a Modular Humanoid for China’s Robot Market. DexForce’s reported story is narrower: a Series B, a simulation-to-real claim, and the need for physical-workplace evidence.
The Independent Workplace Test Still Ahead
The next test is independent evidence in physical workplaces on reliability, cost, safety, and simulation-to-real transfer. This is precise enough to assess the reported claim without asking the evidence to answer a different question. Reliability would show whether the transferred behavior holds up. Cost would test DexForce’s Reuters-reported comparison with repeatedly collecting demonstrations from human trainers. Safety would be separately visible in the workplace evidence. Transfer would examine the passage from training in simulation to behavior on a physical machine.
Each measure answers a separate question. A cost result alone leaves safety unresolved, while simulated preparation alone leaves transfer unresolved. Reported company or customer-use information also awaits independent workplace evidence. What is needed is independent physical-workplace evidence that lets each part of the reported proposition be assessed in its proper place.
For now, Reuters has reported a 1 billion yuan Series B, a valuation above 10 billion yuan, a company led by Jia Kui with roughly 500 employees, and DexForce’s claims about training with simulated data. DexForce’s official pages describe products and a Sim2Real VLA pipeline. The financing is the reported development; the next test is independent evidence on reliability, cost, safety, and simulation-to-real transfer.