Jixian Technology, the "shovel seller" in the edge AI sector, has closed a Series A financing of tens of millions of RMB.

Edge-side AI "shovel seller" Jixian Technology completes tens of millions of yuan Series A financing to accelerate the large-scale implementation of AI hardware

By integrating large models into a toy, a chat-enabled demo can be produced in just a few days. However, when the product enters the mass production phase of 10,000 or 100,000 units, the real challenges emerge: whether the device can clearly hear the user's voice in a noisy environment, respond normally under unstable network conditions, how long the battery can last, and whether the devices can be updated synchronously after model upgrades.

These problems usually do not appear in product demonstrations, but they determine whether an AI hardware product can truly enter the market.

Recently, Shenzhen Jixian Technology Co., Ltd. announced the completion of tens of millions of yuan in Series A financing. This round of financing is jointly invested by Hangzhou Jintou Dingrong Equity Investment Fund Partnership (Limited Partnership) under Zhejiang Provincial Innovation Investment Group Co., Ltd. and Coral Capital. The raised funds will be mainly used for R&D of AI Agent, edge-side AI interaction and embedded development technologies, upgrading of the MarsGate AIoT platform, as well as expansion of AI companion, robotics, AI home appliance businesses and overseas markets.

Founded in 2013, Jixian Technology positions itself as a "shovel seller" in the AI hardware industry, providing brand owners and hardware enterprises with integrated support covering AI capability access, terminal engineering development and large-scale mass production.

AI companion toy product

 The real threshold for AI hardware lies beyond the demo phase

AI is expanding from smartphones and computers to more physical terminals.

In the toy industry, products that originally could only play fixed audios are starting to support free conversation, user information memory and multi-modal interaction; in the home appliance industry, users no longer need to accurately say "turn on the third gear of wind", but can directly express "I feel a little hot", and the device needs to make judgments based on its own status and user habits.

However, integrating large model capabilities into terminals does not mean that the product is ready for mass production.

The iteration speed of models is usually calculated in months, while the development, certification and sales cycle of hardware products may last for several years. Terminals are also restricted by factors such as chip computing power, memory, cost, power consumption, acoustic structure and network environment. A prototype that runs smoothly in the office may face problems such as reduced far-field recognition performance, mutual interference when multiple people speak, slowed response under weak network, audio distortion and insufficient battery life after being placed in real households.

These problems often interact and restrict each other. Increasing local computing can reduce response time, but it may lead to higher power consumption; adjusting the microphone structure can improve far-field recognition, but it may also affect the mold and the overall cost of the device; after the model performance is improved, the hardware resources may not be able to be upgraded synchronously.

"AI hardware is not just adding a chat function to a product," said Wu Kaihua, founder and president of Jixian Technology. "Consumers do not care which model the product invokes, what they finally perceive is whether it can understand them, whether the response is fast enough, whether it often drops offline, and whether it remains stable after being used for several months."

In his view, a new industrial gap is emerging in the AI hardware industry: large model enterprises have algorithm capabilities, but are still far from terminal mass production; traditional hardware enterprises have supply chain and manufacturing capabilities, but lack experience in AI system engineering.

Deploying AI capabilities onto chips and devices

Over the past decade, Jixian Technology has been engaged in smart device connectivity, embedded software and hardware, and IoT platform R&D, and has experienced multiple rounds of technology iterations including Wi-Fi, Bluetooth, Zigbee, cellular communication and Matter. After the rise of large models, the company began to combine these capabilities with AI Agent, voice interaction and long-term memory.

At present, Jixian Technology's solution mainly consists of three parts.

The first part is the AI Agent framework for terminal scenarios, supporting interaction capabilities such as voice, vision, touch and memory. Taking AI companion products as an example, after the user issues an instruction, the system needs to complete voice recognition, intention judgment, content retrieval, model invocation and speech synthesis; if device actions are involved, it also needs to convert natural language instructions into control commands that hardware can execute.

The second part is the embedded development framework. Jixian Technology turns commonly used capabilities such as audio, display, camera, communication, AI recognition, OTA and production line testing into reusable components. For relatively standardized products, part of the code can be generated and compiled through parameter configuration, so that engineers can focus more on hardware adaptation and differentiated functions.

The third part is the MarsGate AIoT full-stack technology platform, which is mainly responsible for device management and operation after shipment. The platform covers links including device access, status management, data analysis, OTA upgrade and fault diagnosis, and supports concurrent management of millions of devices.

MarsGate AIoT full-stack technology platform

For AI hardware enterprises, product shipment does not mean the end of the project. Model invocation, character content, long-term memory, version upgrade and device operation will continue to occur. Therefore, Jixian Technology's business has gradually extended from one-time project development and hardware delivery to software platforms, AI capability invocation and continuous operation services.

Starting from AI companion toys to expand into more terminal scenarios

Up to now, Jixian Technology has served more than 200 brand customers in total, with the cumulative shipment of smart devices and platform access volume exceeding 20 million units. The company currently has about 170 employees, of whom R&D personnel account for about 70%. It has filed more than 70 domestic invention patents in total and owns 54 software copyrights.

In the first half of 2026, the company's overall shipment volume nearly doubled year-on-year, and the revenue from AI-related businesses maintained high growth for four consecutive quarters.

AI companionship is one of the key scenarios that Jixian Technology is currently focusing on implementing. The company has participated in the R&D and mass production of multiple toy and companion products, with relevant capabilities including free conversation, long-term memory, character management, visual recognition and multi-modal interaction, and some products have been launched on domestic and overseas e-commerce channels.

In the field of AI home appliances, Jixian Technology is helping devices shift from executing fixed instructions to understanding user expressions, reading device status and making simple decisions. For example, when a user says "I feel a little hot" to a fan, the system needs to judge whether adjustment is required based on the current gear, environmental status and user habits, instead of mechanically increasing the air volume by one gear.

For home appliance enterprises with a large number of product lines, if they re-develop a set of AI capabilities for each model, the cost and maintenance pressure will rise rapidly. The ability to reuse capabilities across chips and product categories will become the key to the feasibility of platform-based solutions.

State-owned capital and industrial capital jointly bet on AI hardware implementation

This round of financing is jointly invested by Hangzhou Jintou Dingrong Equity Investment Fund Partnership (Limited Partnership) under Zhejiang Provincial Innovation Investment Group Co., Ltd. and Coral Capital.

Relevant person in charge of Zhejiang Innovation Investment said that Jixian Technology has both AI Agent R&D capabilities and hardware mass production delivery capabilities, which can further deploy AI capabilities to terminal devices and real application scenarios.

Coral Capital stated that it will rely on the intelligent manufacturing and supply chain resources in the Yangtze River Delta to help Jixian Technology promote production capacity expansion and industrial collaboration.

After completing this round of financing, Jixian Technology will continue to increase investment in AI Agent, edge-side interaction, embedded development tools and global device operation systems, and expand terminal scenarios such as AI companions, AI home appliances and robots.

Wu Kaihua said, "AI is moving from the cloud to the physical world, and it must finally face the challenges of cost, power consumption, network and mass production. We hope to encapsulate these complex issues, so that our customers can focus more on the products and users themselves."