ThunderSoft bets on robot brains and Agent AIOS, subsidiary Xiaowu Intelligence doubles down on industrial embodied intelligence

Robot Vision reported on October 26 that at the ThunderWorld 2026 Autonomous Intelligence Technology Conference and Product Launch, ThunderSoft (Zhongke Chuangda) unveiled its product and solution lineup across intelligent vehicles, AIoT (smart IoT), embodied intelligence and robotics, and enterprise intelligence.

The offerings showcased included AquaDrive AIOS 3.0, the TurboX autonomous intelligent IoT platform, the Veda AI glasses solution, the WiseEvolve enterprise AI operating system, and the full-stack robotics and embodied intelligence solutions of its subsidiary KNEWBOTS (Xiaowu Intelligence).

Specifically, AquaDrive AIOS 3.0 focuses on the interaction and task execution capabilities of intelligent vehicles; the TurboX platform provides smart devices with support for multimodal perception, task planning, and device invocation; and WiseEvolve, an enterprise AI operating system driven by the physical body, can work in the form of enterprise digital employees to collaboratively execute tasks such as business analysis, customer service, project delivery, and supply chain management. In the robotics field, KNEWBOTS demonstrated system solutions covering robot bodies, control, collaborative scheduling, and data management.

In an interview with the media, ThunderSoft further explained its overall approach to the robotics business. Geng Zengqiang, co-founder and executive president of ThunderSoft, summarized the company's core positioning in the robotics field into two points: first, the "robot brain + Agent AIOS"; second, providing complete robot deployment solutions in key vertical scenarios.

Around these two directions, ThunderSoft is extending its long-term accumulation in operating systems and on-device intelligence into the field of embodied intelligence.

1. Betting on the robot brain and Agent AIOS, ThunderSoft discloses its embodied intelligence layout

ThunderSoft's robotics strategy is strongly connected to its existing businesses.

Since its founding in 2008, the company has gradually expanded its business from smart terminals to fields such as intelligent vehicles and the smart IoT, accumulating capabilities in operating systems, software platforms, on-device intelligence, and system integration. As robots' demand for computing platforms, real-time control, and intelligent systems increases, these technologies have begun to extend into the robotics field.

Geng Zengqiang mentioned that ThunderSoft's most core positioning in the robotics field is the robot brain and Agent AIOS. At the same time, in selected application scenarios, the company will provide complete robot solutions.

This also determines ThunderSoft's approach to investment: different strategies for different industry segments.

Regarding robot bodies, ThunderSoft does not build bodies itself, but mainly establishes cooperation with relevant companies through strategic investment; the group has already invested in Stardust Intelligence.

Regarding models, ThunderSoft will not enter the field to build general-purpose robot brains, but will make strategic investments; the company has invested in two model vendors and is still watching for opportunities in the field of robot world models.

Upstream technologies such as tactile sensors are also within its scope of attention, and the data acquisition devices it produces have already been shipped to a number of domestic robot manufacturers. Some upstream equipment is developed in-house by related businesses under the company, while some capabilities are supplemented through investment and cooperation.

The chip platforms that support the robot brain are one of ThunderSoft's current key investment directions.

Geng Zengqiang introduced that ThunderSoft is carrying out robotics-related layouts around chip platforms such as Nvidia, Qualcomm, Intel, and AMD, with Horizon Robotics as a key domestic partner. Among them, the company has developed a computing system for robots based on the Nvidia Thor platform, in a manner similar to how it develops domain controllers for cockpits in the automotive field.

It is understood that ThunderSoft has reached a cooperation intention with Intel on the establishment of a robotics laboratory, and is also advancing relevant strategic cooperation with AMD.

Geng Zengqiang also talked about ThunderSoft's customer resource advantages in the automotive field. He believes that robots are very similar to cars, and cars themselves are the largest application scenario of embodied intelligence. For automakers, robots are expected to become their second growth point; ThunderSoft has already covered major domestic automaker customers, and in the future will also cooperate with these automaker customers to help them enter the robotics field.

However, Geng Zengqiang maintains a cautious judgment on the pace of development of general embodied intelligence.

He believes that one of the biggest bottlenecks for robots at present is still the insufficient generalization capability of models. Many demonstrations rely on training in specific environments, and if the environment changes slightly, the robot may be unable to complete the task; dexterous hands also have obvious limitations in terms of touch, flexibility, and cost.

Regarding the deployment of robot applications, he judges that at the current stage, closed scenarios and single tasks are more likely to be deployed first, and then gradually cover multi-step tasks in semi-closed scenarios. Commercial scenarios have a higher fault tolerance rate, industrial production lines have stricter requirements for cycle time and stability, and home environments place greater demands on the robot's generalization capability.

This also explains why ThunderSoft chose vertical scenarios as another main business line: at the current stage, customers need deliverable and sustainably operating robot systems, and industry is precisely the key direction of its subsidiary KNEWBOTS.

2. Incubating KNEWBOTS, with one brain for multiple forms and multi-robot collaboration aimed at industrial sites

KNEWBOTS, incubated and controlled by ThunderSoft, mainly targets smart manufacturing and smart logistics scenarios.

At this conference, KNEWBOTS demonstrated a full-stack robotics and embodied intelligence solution, with a product lineup covering forms such as autonomous mobile robots, forklift robots, embodied composite robots, and humanoid robots.

Yang Zhenxiang introduced that KNEWBOTS robots of different forms use a unified controller based on Qualcomm chips and run a unified AIOS. In this way, different models can reuse some underlying software, and it is also convenient to uniformly collect robot operation data.

Robot hardware also adopts a design approach similar to the variable chassis of automobiles. Facing customers' different requirements for size, payload, and drive mode, the team can adjust product configurations more quickly.

In addition, based on RobotOS and multimodal sensor fusion technology, KNEWBOTS builds a closed loop of robot perception, decision-making, and control. Its solution supports millisecond-level real-time control, 3D environment modeling and obstacle avoidance, and draws on an automotive-grade functional safety system to support OTA updates and fault fallback.

But in a real factory, smooth operation of a single robot is not enough.

Yang Zhenxiang explained that the complexity of industrial scenarios is reflected in aspects such as floors, lighting, production line layout, and process flows. Even for the same customer, different factories may have different material categories, work rhythms, and task arrangements. Some production environments also require equipment to maintain stable 7×24-hour operation, and robot task arrangements need to remain consistent with the factory's overall processes.

In reality, factories may also use robots from multiple brands at the same time. Therefore, how to enable different robots to work collaboratively and establish connections with the factory's existing business systems is a key focus of KNEWBOTS' investment.

It is understood that it has built five major platform systems covering scheduling, algorithms, and data, supporting collaborative scheduling of multi-brand, multi-form robots, and through task orchestration, simulation verification, and data feedback mechanisms, promoting the integration of robot capabilities with actual business processes.

In addition to on-site collaboration, KNEWBOTS is also building an industrial data closed loop.

Its unified controller can provide a foundation for data collection from different models, and the supporting data platform is responsible for data collection, cleaning, labeling, and structured processing. The relevant data can be used for subsequent model training, and the optimized models are then deployed to robots via OTA.

Considering the data security requirements of industrial customers, Yang Zhenxiang mentioned that for customers requiring data not to leave the factory, KNEWBOTS will deploy edge-side products at the factory site, so that data processing flows on the device side, edge side, and related cloud side all remain within the factory.

In addition, Yang Zhenxiang revealed that the team is also advancing products and overall solutions in directions such as one-click deployment and out-of-the-box use, hoping to further lower the threshold for deploying and using industrial robots.

Conclusion: Robot deployment must ultimately be proven in real scenarios

For ThunderSoft, robotics and embodied intelligence are becoming important application directions for its operating systems, edge intelligence, and engineering capabilities. Its long-accumulated underlying technologies provide a foundation for expanding its robotics business, while KNEWBOTS further undertakes product R&D, system integration, and application deployment in industrial scenarios.

As embodied intelligence gradually moves toward real scenarios, the industry's focus is also beginning to shift from single-machine capability demonstrations to system collaboration and practical applications. Whether robots can stably complete tasks, adapt to complex environments, and deeply integrate with customers' existing production and business processes is becoming the key to promoting industrialization.

From ThunderSoft's layout, it can be seen that the industrialization of embodied intelligence requires the collaborative support of model algorithms, operating systems, robot bodies, and engineering capabilities. In the future, as robots gradually integrate into more actual business scenarios, how to transform technological capabilities into stable, easy-to-use, and sustainably operating solutions will become an important issue for the embodied intelligence industry as it moves toward large-scale application.