Gasgoo Munich- On April 27 that Anyverse Dynamics, a general embodied intelligence robotics company, has closed its Angel++ financing round.
Meanwhile, the Angel+++ round is nearing completion. It is backed by internet industrial capital and top-tier U.S. dollar and RMB funds. Total financing in the angel series has now surpassed $200 million.
Proceeds will be used to accelerate the development of a general embodied brain based on the "Native World Model + Reinforcement Learning" paradigm. The funds will also build technical infrastructure such as efficient data pipelines.
Anyverse Dynamics also announced that it has signed with Envision Group a global market order worth over 500 million yuan. It targets deployment across Europe, Asia, and other regions worldwide.
Image Source: Anyverse Dynamics
Unlike the industry-standard Vision-Language-Action (VLA) architecture, Anyverse Dynamics has centered its approach on a latent space world model. Combined with reinforcement learning, the company is building MWA™. It is a fully pre-trained, general embodied brain native to the physical world.
MWA™ unifies the representation, training, and inference of multimodal information within a latent space. This approach overcomes the shallow limitations of traditional visual perception. It enables the embodied brain to deconstruct the underlying logic of the real world. It also grasps objective physical laws and causal relationships. It fundamentally avoids information attenuation and feature fragmentation caused by cross-representation mapping in the VLA paradigm. Furthermore, it discards computational overhead and excessive noise associated with redundant pixel-level predictions in World Model architectures like WAM/VAM. Instead, MWA™ employs a lightweight, energy-efficient algorithmic architecture. This equips robots with core capabilities for physical world cognition and future state prediction. Reinforcement learning, meanwhile, continuously iterates and optimizes decision strategies through autonomous interaction with dynamic environments. It builds a value decision system that is complete and aligned with real-world logic. The two components are deeply coupled and operate in a closed-loop synergy. They drive the embodied brain to achieve a leap in capability and exponential evolution in intelligence.
Co-founder and Co-CTO Xia Zhongpu is one of China's earliest researchers in world models and reinforcement learning. He brings over 15 years of deep experience. As the creator of China's first end-to-end autonomous driving system, he possesses rare, systematic engineering capabilities for complex AI hardware and software systems. Xia and his team will leverage this expertise and their capacity for continuous innovation. They will accelerate the R&D and application of native world model and reinforcement learning technologies in embodied intelligence.
About Seeds Discovery:
Gasgoo's "Seeds Discovery" column creates a service platform connecting startups, industry ecosystem partners, investors, and local governments. It provides deep empowerment across the supply chain. Since its inception, the column has been dedicated to identifying high-potential companies, technologies, and business models. These offer inspiration and leadership during the wave of intelligent transformation. This fosters the growth of innovative forces in the automotive industry. According to Gasgoo statistics, nearly every startup featured in "Seeds Discovery" has successfully integrated with industry ecosystem resources.
