Beijing’s First Embodied Data Unicorn Secures 1 Billion Yuan in Funding

1 billion yuan financing has led to the emergence of the world’s first embodied data unicorn in Beijing. This milestone was reported on March 21, 2026, by the Beijing Daily Client, highlighting significant developments in the field of artificial intelligence and robotics.

During the Spring Festival gala, multiple humanoid robots made a collective appearance, bringing embodied intelligence into the public spotlight. As humanoid robots accelerate towards mass production and integrate into factories and daily scenarios, a pressing question arises: do these robots require “feeding” to operate? The answer is yes; however, the “feeding” consists of high-quality, vast amounts of data, supported by advanced simulation technologies. This need has made data-providing companies the darlings of the industry.

Recently, Lightwheel Technology completed a financing round of 1 billion yuan, comprising A++ and A+++ rounds. This round attracted several industry players and financial institutions, including New Hope Group, Dingbang Investment (the family office of the chairman of Sanan Optoelectronics), AUX, and Dingshi Asset Management. State-owned institutions such as China Investment Huake and Guofang Innovation also participated, alongside financial investors like Daohua Long-term Investment and Qingxin Capital. Lightwheel Technology plans to focus on key areas such as the continuous research and development of physical simulation engines, upgrading large-scale model evaluation systems, and enhancing delivery and deployment capabilities, further solidifying its position in the field of physical AI data and simulation infrastructure.

In recent years, Beijing has accelerated its focus on artificial intelligence and robotics. The development trajectory of Lightwheel Technology illustrates that the infrastructure for embodied intelligent data is transitioning from cutting-edge technology exploration to large-scale industrial applications, becoming another pillar of technological innovation in the capital.

2026 marks the beginning of the era of scaled embodied data. As the quantity of robots increases, the variety of application scenarios expands, and the task chains lengthen, the data requirements grow significantly. This growth is not linear; it resembles an exponential leap. Once large-scale implementation is initiated, the data gap will widen geometrically.

Similar to how athletes require repetitive training to master complex movements, every precise grasp and agile operation performed by robots necessitates a vast amount of data “feeding.” Since the second half of 2025, industry practices have confirmed that the key to enhancing robotic capabilities is not merely repetitive practice of the same action but rather exposing the robots to diverse environments on a larger scale. However, the collection of real-world data presents inherent challenges, including high costs, low efficiency, and the inability to replicate scenarios. The speed of data supply is far outpaced by the demand for model iteration. Simulation is emerging as the crucial key to overcoming this bottleneck.

From a policy perspective, the Ministry of Industry and Information Technology’s Guiding Opinions on the Innovative Development of Humanoid Robots clearly states the need to “establish simulation systems and training environments.” The 14th Five-Year Plan emphasizes improving the self-controllable level of the industrial chain, and embodied intelligence has been identified as a future industry in the plan’s forward-looking layout. The industry consensus is becoming increasingly clear: simulation is not merely an auxiliary tool; it is a prerequisite for effective robotic learning.

With the arrival of 2026, the era of scaled embodied data is officially here.

Three-tier architecture and all-stack self-developed simulation core: Lightwheel Technology has established a comprehensive technical system covering simulation generation, large-scale data production, and model capability evaluation based on a “simulation—data—evaluation” three-tier architecture.

At the simulation layer, Lightwheel Technology has developed a world-first all-stack self-developed technology architecture that integrates “solving—measuring—generating.” Its proprietary solver supports high-precision real-time solving across rigid bodies, soft bodies, and fluid dynamics, combined with an innovative physical measurement factory and virtual-real alignment methodology. This allows real-world physical parameters to be accurately transferred into the digital space, making simulation not just visible but genuinely applicable for training. This simulation core has been deeply adapted for domestic chips, ensuring self-control and effective compatibility with the domestic computing power ecosystem.

At the data layer, Lightwheel Technology has built the world’s largest non-entity data engine, covering both simulation-generated data and human video data, enabling the large-scale production of embodied data. Lightwheel’s data solutions deliver high-frequency, scalable data fuel for foundational models.

At the evaluation layer, Lightwheel Technology has launched the industry’s first industrial-grade simulation evaluation platform, RoboFinals, which is based on 100 challenging tasks to create a standardized evaluation system. Alibaba’s Tongyi Qianwen has already partnered with Lightwheel on this platform to advance the establishment of industrial-grade evaluation standards, forming a closed-loop mechanism of “problem discovery—targeted learning—continuous improvement.”

The three layers of capabilities support and reinforce each other: the simulation layer strengthens the foundation of physical reality and environment, the data layer drives large-scale generation and ongoing optimization, and the evaluation layer provides systematic feedback that propels iterative capability improvements. As these three layers continuously interact in a closed loop, Lightwheel Technology’s simulation accuracy, data quality, and evaluation depth consistently improve, enhancing the overall infrastructure capabilities.

10x growth, opening the treasure trove of industrial data: With this comprehensive capability system, Lightwheel Technology has achieved large-scale delivery in simulation-generated data, simulation evaluation, and first-person human video data, leading in all three delivery dimensions.

Recently, Lightwheel Technology was featured in a special report by CCTV, where co-founder and president Yang Haibo stated that the quality of simulation technology and data is ultimately judged by the customer. He emphasized that Lightwheel not only delivers data but also possesses algorithmic capabilities, allowing collaboration with clients to iterate on the “recipe” for data. Currently, leading enterprises such as ZhiYuan Robotics, Galaxy General, ByteDance, and Alibaba are among Lightwheel’s clients or partners. With its all-stack self-developed simulation technology, Lightwheel has quickly gained market recognition, and its commercial progress is accelerating. Reports indicate that Lightwheel achieved a tenfold revenue increase in 2025, with projected Q1 revenue for 2026 already surpassing the entire revenue of 2025.

Yang Haibo stated, “Our commitment to all-stack self-developed simulation stems from the current simulation software’s inability to meet our needs. We must ensure our data generation and evaluation capabilities through comprehensive self-development. Additionally, we recognize the critical supporting role of simulation as a foundational technology for humanoid robot development.”

As embodied intelligence transitions into practical applications, more industrial companies are beginning to focus on the sources of robot training data. In this financing round, manufacturing giants like New Hope Group, AUX Group, and Sanan Optoelectronics joined forces, covering scenarios such as food processing and home appliance assembly, not only as investors but also as actual users of Lightwheel’s technology and partners in industrial collaboration.

Through first-person video capture, physical measurement, and simulation modeling, the operational experience and processes accumulated over the long term in factories are being transformed into data resources that robots can learn from. Lightwheel Technology plans to focus on three directions: building first-person human video data collection capabilities with industry partners; deepening collaborative physical measurement to accurately input real production line physical parameters into the simulation world; and promoting deployment validation from simulation to reality, ensuring that skills trained in simulation can be effectively applied on production lines.

In fact, the value of industrial data extends far beyond manufacturing. Scenarios such as inspection and maintenance of urban rail transit, logistics scheduling in large commercial complexes, and intelligent operations of municipal infrastructure also harbor vast amounts of operational experience that remain undigitized. Lightwheel’s all-stack self-developed simulation core and first-person data collection capabilities are providing a complete transformation pathway from physical measurement to data production for these complex scenarios.

As more industrial enterprises engage in collaboration, Beijing, as a key hub for artificial intelligence, rail transit, and smart city industries, is poised to form a synergistic ecosystem of “industrial scenarios + embodied data.” Lightwheel’s network of industrial cooperation is also expanding from manufacturing to broader scenarios such as urban operations and transportation logistics. Real industrial scenarios are gradually becoming the most crucial data source for embodied intelligence, and Lightwheel Technology is emerging as a vital bridge connecting industrial scenarios with robot training through its comprehensive capabilities in simulation modeling, physical measurement, and data production. Cities and industries with rich scenarios are becoming strategic “data sources” in the era of embodied intelligence.

The era of physical AI’s data and simulation infrastructure: Numerous institutions predict that in 2026, leading humanoid robot manufacturers will achieve deliveries in the thousands, with annual sales expected to exceed 100,000 units. Surpassing this threshold will again elevate the demand for training data and evaluation capabilities by an order of magnitude.

Lightwheel Technology’s goal has never been to produce a specific robot but to establish the infrastructure for physical AI data and simulation. Just as GPUs defined the computational infrastructure for large models, the infrastructure composed of simulation worlds, behavioral data, and evaluation systems is equally indispensable in the physical AI era. Those who control this infrastructure will hold the initiative in industry iterations.

The year of the large-scale explosion of embodied data has arrived, and Lightwheel Technology is deeply rooted at the core of this transformation. Originating from Beijing, leveraging its all-stack self-developed and self-controlled physical AI simulation technology, Lightwheel is accelerating the transition of embodied intelligence from the laboratory to various industries, paving the way for humanoid robots to truly enter the real world and solidifying the technological foundation.