DexForce Closes RMB 1 Billion Series B

Recently, DexForce Technology, a leading Physical AI and embodied AI company, officially announced the completion of a Series B round of RMB 1 billion, reaching a post-investment valuation of more than RMB 10 billion, becoming an embodied AI unicorn valued at RMB 10 billion, and moving quickly toward an IPO.

This round brings together a formidable lineup of investors, with national-level funds of funds, leading state-backed venture capital firms, industrial capital from real-economy champions, and local sci-tech innovation platforms all converging. Among them, Shenzhen Capital Group and the Guiyang Digital Economy Fund have firmly doubled down with heavyweight investments in two consecutive rounds; the Qianhai Mother Fund, Lens Technology, ICBC Capital, Hengjian Assets, and Zhurui Capital joined as powerful new investors; Nanshan Strategic Emerging Industry Investment, Chengdu Science and Technology Innovation Investment, and the Sichuan Academician Fund and other existing shareholders continued to add follow-on investments. With new and existing investors alike doubling down, this round is a resounding vote of confidence from the capital market in the long-term technology roadmap, real-world deployment capability, and long-term industrial growth value of DexForce Technology.

The new funding will be focused on iterating core world model algorithms, upgrading the physics engine, building data infrastructure, refining humanoid robot capabilities, and landing real-world scenarios, continuously accelerating the co-evolution of technology, products, and the business closed loop.

Since its founding in 2021, DexForce Technology has been anchored in the core tracks of Physical AI and world models, completing its technology layout with long-termism and standing as one of the earliest benchmark companies in China devoted to fully in-house Physical AI. DexForce firmly chose world models + physical simulation + real-robot deployment as the end-game technology path for the industry. Its research cadence is strikingly forward-looking, with progress on multiple core technology milestones substantially ahead of overseas giants such as NVIDIA and DeepMind, and its route choices over the years have kept being validated by global industry trends.

Recently, DexForce DexWorldModel topped the leaderboard of WorldArena, the authoritative global benchmark for embodied world models, with its technical strength authoritatively validated by the industry. As the core self-developed embodied world model of DexForce, DexWorldModel is built around the true operating laws of the 3D physical world, possesses latent-space dynamic modeling capability, and achieves unified coordinate-system alignment and precise spatiotemporal representation matching. It thoroughly resolves such industry pain points of traditional vision solutions as inaccurate spatial perception, missing physics modeling, and inability to support hands-on tasks, truly enabling AI to accurately understand physical space, dynamics logic, and real operating rules, and supporting robots in practical tasks such as sorting, assembly, and dual-arm composite operations, moving the world model officially from the stage of “visual demonstration” into the industrial stage of “deployable real-world work”.

While consolidating model capability, DexForce Technology has also gone deep into data infrastructure, innovatively proposing Dexterity-BEV, building benchmark-grade data infrastructure for the field of Physical AI. This technology aligns visual input, robot embodiment state, and target action commands onto one unified spatial reference frame, for the first time giving multi-source heterogeneous robot data a standard spatial foundation that supports training at scale. It eradicates from the root the long-standing pain points of scattered data standards, manipulation skills that hardly transfer, and operating experience that cannot be reused across embodiments, greatly improving data utilization efficiency and cross-scenario model generalization, and laying a solid data-infrastructure foundation for the large-scale industrial deployment of Physical AI.

DexForce Technology adheres to first principles, holding that the essence of general embodied AI is generalization within scenario boundaries. Through precise definition of scenario boundaries, it adapts data, models, and training paradigms to product-grade operating standards, achieving high-precision, high-reliability general capability within each scenario, and keeps widening capability boundaries through scenario-by-scenario technical breakthroughs, steadily advancing toward the ultimate Physical AGI.

Building on four self-developed foundational platforms — the DexVerse™ physics engine, the DexAgent general embodied AI agent, the DexSense spatial perception terminal, and the DexBrain intelligent algorithm brain — DexForce Technology has established a complete integrated software-hardware capability foundation that supports robust deployment across all scenarios.

To date, DexForce Technology has deployed 50+ diversified application scenarios and iterated and accumulated more than 1,500 mature embodied models, covering segments such as automotive components, new energy, 3C electronics, aerospace, logistics, home appliances, chemicals, healthcare, and education. Serving many industry-leading customers, it brings mature enablement to flexible sorting, flexible insertion-and-assembly, commercial retail, and cultural and tourism services in intelligent manufacturing, standing as the leader in large-scale commercial deployment of embodied AI in China.

DexForce Technology was also the first to propose Physical Token Economics, redefining the development logic of the Physical AI industry. Controlling cost and efficiency across the whole chain from data collection, model training, inference and deployment, and hardware embodiment to scenario operations and maintenance, DexForce focuses on optimizing the input-output ratio of each single piece of data, each single action, and each single inference, pursuing the highest commercial return on every technical capability and every robot operation, and driving a deep match between technological innovation and business value.

Prof. Kui Jia, founder of DexForce Technology, said: “Our core pursuit is to achieve the highest ROI per physical token — so that in the physical world, every set of data, every action, and every round of inference converts efficiently into real commercial value. DexForce has always been rooted in underlying data and models, building Physical AI capabilities that generalize, deploy, and commercialize, and continuously validating and closing the value loop across industrial and commercial settings as well as the gradually maturing elderly-care and home scenarios.”

Proven technology deployment and a working business loop are the core logic behind this round of capital doubling down on DexForce Technology. After years of deep cultivation, DexForce Technology has formed “technology iteration — product deployment — customer payment — data feedback” as a positive business cycle, becoming one of the few companies in the industry to achieve revenue exceeding RMB 100 million and large-scale shipments of robots performing real-world work. Notably, Lens Technology, which made a strategic entry in this round, will pursue deep industrial synergy with DexForce, focusing on landing core intelligent-manufacturing scenarios and fully accelerating the large-scale commercialization of Physical AI technology across the real manufacturing sector.

Amid the heat of the track, DexForce Technology has maintained strategic composure, taking “advancing toward general intelligence for the physical world” as its mission and following “delivering results along the way, deploying step by step” as a steady development path, and building sustainable core industry barriers. Reinforced by the funding and industrial resources of this round, DexForce Technology will keep strengthening its three core moats — technology, mass production, and commercialization — steadily advancing the industrial deployment of Physical AGI and bringing hard-core self-developed technology to physical intelligence across industry after industry.

An investor from Qianhai Fangzhou said: “DexForce is one of the outstanding companies Qianhai Fangzhou has backed in the embodied AI track. Relying on its fully self-developed simulation engine, it can generate synthetic data efficiently at scale and use a “world model” brain to guide humanoid robots working on site in retail, tour guiding, industry, and other scenarios. Its world model (DexWorldModel) also took first place on the WorldArena Track2 leaderboard; the model accurately predicts how physical interactions change in the real 3D world (rather than merely generating videos), fully demonstrating the technical strength and deployment capability of the team core to Prof. Kui Jia. We expect full-year revenue to reach several hundred million yuan this year, remain confident about the future of the company, and look forward to its smooth entry into the capital market.”

An investor from Lens Technology said: “As a core industrial investor in DexForce Technology, we remain confident in the long-term embodied AI track driven by Sim2Real virtual-real synergy. DexForce Technology follows the pragmatic route of “learning the laws of physics in the virtual world first, then landing in real industrial scenarios”. With its fully self-developed DexVerse™ generative simulation engine at the core, it has built a complete four-layer product matrix spanning the simulation platform, spatial perception sensors, the embodied intelligence brain, and complete humanoid robots, and through a coordinated pipeline of real data and synthetic simulation data it sharply lowers robot training costs and lifts scenario generalization, striking at the core pain points of deployment in manufacturing and commercial services. Lens Technology has specialized in precision intelligent manufacturing and volume production of humanoid robot hardware for many years. Going forward, we will leverage our full-industrial-chain manufacturing, large-scale delivery, and scenario resources to synergize deeply with DexForce, connecting simulation algorithms, perception software, and the mass-production chain of complete robots, jointly moving embodied AI technology from the laboratory to large-scale industrial deployment, and powering the intelligent upgrade of the real manufacturing sector.”