LimX Dynamics, Having Raised $400 Million in Half a Year, Accelerates IPO Timeline

The Embodied AI track's multi-billion-dollar unicorn adds another powerhouse.

On 7 April 14, general-purpose humanoid robot company Piaggio Motion announced the completion of its Pre-IPO financing round, raising nearly 200000000 US dollars and achieving a post-money valuation of 15000000000 yuan. Over the past half year, Piaggio Motion has cumulatively raised 400000000 US dollars in financing.

This also means that the IPO process for ZJDL has begun.

Observing the investors in this financing round reveals a collection of top-tier international investment institutions from around the world. These include: IDG Capital, Lens Technology, GGG Group, Redstone VC, Huashan Capital, and Hefei Binhu Industrial Development Group. Additionally, UAE-based Stone Venture has continuously invested in multiple rounds, providing significant support; existing shareholders such as Oasis Capital, Cornerstone Capital, Nanshan Strategic Emerging Investment, Shangqi Capital, and NIO Capital have all over-subscribed their allocations.

This financing round gathers industrial capital, international capital, and long-term capital from China, Europe, the Middle East, and North America. This represents the capital market's strong confidence in the Embodied AI track, as well as high recognition of ZJDL's technological capabilities, product capabilities, and global strategy.

According to plans, the funds from this round will be primarily used to break through and commercialize brain-body fusion technologies, promote the large-scale deployment of thousands of fully autonomous humanoid robots, and strengthen global market expansion. Meanwhile, ZJDL will further improve its global manufacturing and delivery capabilities, accelerate layout in key markets such as the Middle East, Europe, and other Asian countries, and continue to build an open ecosystem for global developers, accelerating the innovation and industrialization of Embodied AI technologies.

Across the entire Embodied AI industry, most companies are still in their early startup stages. No more than 13 have entered the ranks of billion-dollar unicorn status, and those rushing toward an IPO are as rare as phoenix feathers. So why can LimX Dynamics (Zhujidongli) be the first to enter the IPO process?

Behind this lies a close connection to LimX Dynamics' product and technological core. LimX Dynamics was the first to propose and layout the 'Humanoid Robot Brain System,' continuously investing in R&D around the robot's本体 (body), AI infrastructure, and the Embodied AI Agent OS system. This has formed a three-layer technical architecture covering: System 0 full-body motion foundation models, System 1 humanoid robot VLA/WAM capabilities, and System 2 the Embodied AI Agent Operating System (COSA), which uses large language models and world models as its engine.

The world's first Embodied AI agent system, COSA, has achieved a fully autonomous process for humanoid robots to perceive the world, understand tasks, decompose/plan, and interactively execute. Its real-world demonstration time even preceded that announced by US humanoid robot company Figure for Helix 02.

On the product front, LimX Dynamics' robot products have already secured thousands of orders, with more than half coming from overseas markets. The full-size interactive humanoid robot LimX Luna, released in month 5 this year, had already begun bulk deliveries to domestic and international markets by month 6. Additionally, the new generation multi-form Embodied robot TRON 2 has also received numerous domestic and international orders.

IPO Process is Accelerating

A deep dive into LimX Dynamics' financing history reveals that this general-purpose humanoid robot company, founded in year 2022, has completed seven rounds of financing. According to public information, the total financing amount has exceeded 3500000000 yuan.

Entering the Pre-IPO funding round, LimX Dynamics introduced more global capital and simultaneously announced its valuation and progress on the IPO for the first time. This funding round serves as a critical springboard for LimX Dynamics moving from the primary private market to the secondary public market.

For investment institutions, early A and B round investments were essentially 'bets' on whether the company could survive. However, the Pre-IPO round is quite different. By this stage, the company's business model is basically proven, and its market position is relatively stable. Investment institutions place greater emphasis on the steady returns that may come after the IPO.

Actually, in 3 this year, ZhiJi Dynamics completed its shareholding reform and was officially renamed "Shenzhen ZhiJi Dynamics Technology Co., Ltd.", obtaining the "admission ticket" for an IPO. Combined with the Pre-IPO financing round, ZhiJi Dynamics' IPO process is accelerating.

Three-Layer Technical Architecture Building a Full-Stack Moat

In 2025, the global production volume of humanoid robots was approximately 20000 units, with almost no true落地 (landing/commercialization) applications capable of performing actual work. In 2026, humanoid robot production is expected to exceed 100000 units. As humanoid robots enter a stage of scaled development, they are also rapidly moving into commercial application and implementation phases.

At this stage, whoever masters the generalization ability of models in the real world will seize the opportunity for implementation first. The core card that has enabled ZhiJi Dynamics to continuously attract heavy investments from top-tier global long-term capital is its three-layer full-stack technical architecture centered around the "humanoid robot brain system": System 0, System 1, and System 2.

Among these, System 2 (Embodied AI Agent OS - COSA), serving as the top-level hub, uses large language models and world models as dual engines. It constructs a complete loop of human-like advanced cognition through four core modules:

• World Model (Understanding and Prediction): This is the key for robots to understand physical common sense. It endows robots with deep perception of space, geometry, and physical laws, allowing them not only to see objects in front of them but also to predict the consequences of every action they take in the physical world (e.g., if a cup falls, water will spill). This enables the deduction of "intuitive physics" in virtual environments.
• Thinking and Reasoning (Closed-loop Reasoning and Action): Moving away from rigid, single-program instructions in the past, the system possesses long-chain task planning (Task Planning) and logical reflection capabilities similar to humans. Faced with vague, abstract, long-cycle tasks like "bring me a cold Coke," the system can autonomously break down and reason through multiple steps; more importantly, when unexpected interference in the real physical environment causes an action to fail, it performs real-time closed-loop error correction and reflection, adaptively adjusting the next strategy.
• Hybrid Memory System (Experience Accumulation): Robots are no longer machines that "lose memory every time they boot up," but evolving entities with fused short-term and long-term memory. Short-term memory locks onto the current stream of environmental states, while long-term memory沉淀 (accumulates/precipitates) immersive semantic knowledge and historical operational experience, making the robot smarter the more it is deployed and explores.
• Diverse Human-Machine Interaction Capabilities: Breaking through multimodal interaction boundaries, it can accurately understand complex natural language logic from humans and capture implicit needs from the environmental context during interaction, achieving truly empathetic and highly cooperative natural collaboration.

However, a smart System 2, lacking powerful execution pathways, would become a "giant in thought, dwarf in action." ZhiJi Dynamics' strength lies in its seamless integration of the big brain and small brain.

Looking downward, the System 1 (i.e., humanoid robot VLA/WAM capabilities) leverages the LimX VGM/DreamActor training paradigm to precisely translate the decisions of System 2 into real-time motion control strategies for the entire body. Meanwhile, FluxVLA Engine, an enterprise-grade open-source infrastructure, provides global developers with an engineering foundation for continuous model training, iteration, and deployment. It frees the industry from the underlying engineering burden of 'reinventing the wheel,' driving the sector forward from 'training a single good model' to 'enabling everyone to train models.' The excellent embodied AI models gathered within this open-source framework will serve as an inexhaustible source for the growth of robotic skills.

At the lowest level, the System 0 (whole-body motion control layer) integrates LimX's motion foundation model with LimX Mimic's human-like action imitation capabilities. This ensures that no matter how complex the decisions made by the high-level brain are, the robot's physical hardware can deliver absolutely stable, precise, and agile anthropomorphic dynamic feedback in unstructured terrains.

Based on this three-tier technical architecture, LimX Dynamics possesses the confidence to compete alongside top international enterprises like Figure. With the quiet completion of its joint-stock reform in month 3 this year, this super unicorn—already holding a 'ticket' to the secondary market and fully armed with a humanoid brain system—is accelerating at full speed to bring general-purpose humanoid robots out of laboratories and truly into diverse industries across the physical world.