Former Horizon Robotics executives start a robotics foundational intelligence venture, securing tens of millions of yuan in seed funding.

```

On May 19, news broke that the space intelligence large-model startup "DingDang Power" announced the completion of a tens of millions seed round of financing.

This round was led by Horizon Robotics, with Zhengjing Fund following. It is understood that the funding will mainly be used for the development and validation of the "space intelligence large model + physical Agent" framework, as well as the construction of a real scenario data closed-loop system. As embodied intelligence investment becomes more rational, this financing reflects industrial capital’s phased preference for teams with comprehensive experience from algorithm fundamentals to hardware implementation.

Public information shows that DingDang Power was established only a few months ago, and its core leader is Niu Jianwei, a former executive of Horizon Robotics. Reviewing his professional background, he is an industry veteran who has grown up with China’s deep learning wave.

In his early years, Niu Jianwei worked in Baidu's Voice Technology Department, participating in early GPU-based deep learning model training at Baidu IDL in 2012. In 2015, invited by Yu Kai, he joined Horizon Robotics during its early startup phase, rising from Algorithm Engineer to General Manager of the Intelligent Cockpit Product Line.

This complete experience spanning algorithm construction, chip adaptation, and eventual product mass production is rare among current large-model startups. Most past AI entrepreneurs have focused on the algorithm layer, but in the real physical world, AI implementation is heavily constrained by computing power platforms, sensor accuracy, and system engineering capability.

Previously, in 2023, Niu Jianwei proposed the concept of low-cost data stimulation through vertical domain post-training, demonstrating his long-term focus on the core issues of "cost and scale" in AI commercialization. This also laid the groundwork for his later decision to enter the embodied intelligence field from the underlying architecture.

According to disclosures, DingDang Power’s core business focuses on “space intelligence large model + physical Agent.” To understand the essence of this business, it needs to be viewed in the context of the current evolutionary path of robotics technology.

Over the past decades, traditional robotics has mainly relied on cybernetics thinking, that is, writing rules and setting parameters for single tasks, with very poor scenario generalization capability, resulting in high customization costs.

As large models exploded, the industry began to attempt joint training of vision, language, and actions, forming VLA models. Yet, current VLA models face obvious practical shortcomings in engineering: the difficulty of multimodal data alignment, limited model scalability, and inability to achieve low-cost continuous self-evolution in the physical world.

DingDang Power’s technical path seeks to avoid this blind spot. Its solution is not merely fine-tuning a single algorithm, but building a system-level solution—allowing the large space intelligence model to understand complex physical environments, while the physical Agent serves as the execution layer, deeply integrating with physical entities.

The essence of this architecture is an attempt to bridge the engineering gap for general large models to output intelligence from “digital screens” to “physical entities.”

In today’s capital environment, financing in the large-model track is concentrating on targets with clear business models and engineering implementation capabilities. DingDang Power received the lead investment from Horizon Robotics early in its founding for two core reasons:

First, deep strategic synergy value. Horizon Robotics, as a provider of underlying intelligent computing platforms, has a core strategy not limited to intelligent driving but also needs to extend its software-hardware integration capability to broader robotics and embodied intelligence ecosystems.

DingDang Power’s exploration at the physical Agent layer can be seen as Horizon’s computing ecosystem naturally extending into physical entities.

Second, the ability to build mass-production data closed-loop systems. The financing specifically states funds will be used for “building a real scenario data closed-loop system.” In embodied intelligence, the core barrier is who can obtain the highest-quality physical world interaction data at the lowest cost.

Niu Jianwei’s practical experience in smart cockpits gives him deep understanding of the flywheel effect of mass-production data feedback and model iteration. Compared to purely academic teams, this systematic capability to build industrial-grade data cleansing, annotation, and training closed loops is a foundational asset valued highly by industrial capital.

Looking ahead, the main battlefield of AI is shifting from pure digital content generation to entity intelligence with strong interaction with the physical world. The “space intelligence” track where DingDang Power operates is the core infrastructure of this transformation period.

However, this remains a long, snowy slope full of unknowns. The long-tail scenarios of the physical world are extremely complex—lighting, friction, gravity, and other common physical knowledge are still huge cognitive gaps for current large models.

DingDang Power will inevitably face substantial data collection costs and engineering challenges of adapting to multi-end hardware as it builds high-quality data closed loops in the short term.

For continued observation of this company, the key is whether it can, within one to two years, deliver a benchmark case of physical intelligent agents in real commercial scenarios that possess a certain degree of generalization capability and controllable costs.

Risk Warning and DisclaimerThe market has risks, and investment requires caution. This article does not constitute individual investment advice, nor does it take into account the specific investment objectives, financial circumstances, or needs of individual users. Users should consider whether any opinions, views, or conclusions expressed herein suit their particular situation. Invest accordingly at your own risk. ```