Red Bear AI schließt Serie-A-Finanzierung in Höhe von 210 Millionen Yuan ab und betritt den Bereich Physik-AI | Exklusivbericht von 36Kr

36Kr has learned that the enterprise AI solutions provider "Red Bear AI" recently completed a Series A financing of 210 million yuan. The post - financing enterprise value is over 1.5 billion yuan. This round was led by Huayu Venture Capital, and the old shareholders Greifeng Investment and Jiawo Capital also participated in the investment. Several institutions such as Xuhui Capital, SAIC US Dollar Fund, Jiaming Haochun and Yuhua Asset also participated in the investment. Three founding team members also participated in the financing with about 30 million yuan based on the current enterprise value. This is already the fourth consecutive round in which Huayu Venture Capital has invested in Red Bear AI.

Red Bear AI, founded on April 7, 2024, focuses on the technology of "Memory Science + Full - Modal Large Model". The company aims to endow AI with a "episodic memory" ability similar to that of humans, so as to enable more precise perception, decision - making and action in the physical world. Founder and CEO Wendeliang told 36Kr that Red Bear AI is expected to apply for listing on the Hong Kong Stock Exchange in the second or third quarter of 2027 and officially start trading in 2028.

From "Product Thinking" to "Technology - Ecosystem Thinking"

"What is your core competence?" This is the most frequently asked question by investors to Red Bear AI during the financing process.

Wendeliang's answer is a bit unusual: "Products are never the core competence. It's like the relationship between a goose and an egg: The egg is a fixed result, but the goose that continuously lays eggs is the real value driver."

In Wendeliang's view, the "ecosystem" of Red Bear AI is the technology based on "Memory Science + Full - Modal Large Model". Memory science solves the problem of how AI can "reuse experiences", and the full - modal large model solves the problem of how AI can "perceive the world". The combination of the two enables the continuous development of innovative applications tailored to different scenarios. Wendeliang believes that products are never the core competence, but the underlying technology system that supports product development (i.e., the process of "raising the goose") is the value driver.

This idea is based on an insight into the problems in the AI industry. A study by Google DeepMind shows that even the most advanced language models can get into "reversal courses" - after learning that "Plato taught Aristotle", they can't answer "who Aristotle's teacher was". The reason is that AI can only learn fragmented knowledge about parameters but is unable to establish a connection between experiences and scenarios like humans.

Red Bear AI has digitized the mechanisms of human "episodic memory" and built an "Intelligent Memory System" similar to the hippocampus. At the same time, it integrates multimodal information such as text, images, speech and sensor data, so that AI can not only "learn knowledge" but also "remember experiences and habits".

"We have developed our current application products in the form of software results. After listing, we want to pursue a second growth curve with hardware products in memory science." Wendeliang revealed, "Currently, the focus is still on the sales of applications."

In contrast to many To - B companies that focus heavily on sales and AI companies that focus heavily on technology but both pay relatively little attention to brand building, Red Bear AI has already started to pay attention to brand building in the early stage.

"We have two core elements in our external communication: First, we win through technology, and second, we implement technology as a brand strategy." Wendeliang shared, "In this way, it is more effective than discussing business models and industry models, and customers also accept it better."

Founder and CEO Wendeliang

Entering the Physical AI Industry: From Fragmented Perception to Digitalization of "Episodic Memory"

In the wave of "Physical AI" defined by NVIDIA CEO Jensen Huang, the ultimate mission of AI is to break through virtual boundaries and become an "Intelligent Agent" in the physical world. However, current AI systems generally encounter bottlenecks of "less accurate perception, less flexible decision - making and unfeasible actions" when entering the physical world.

The core technology of Red Bear AI draws inspiration from cognitive science. The human hippocampus has the ability of "episodic memory", which can store complete scenarios and quickly reuse them. In contrast, traditional AI systems can often only learn fragmented knowledge about parameters and lack the ability of "latent learning" to establish a connection between experiences and scenarios.

The technological breakthrough of Red Bear AI lies in the digitalization of human "episodic memory" and the construction of an "Intelligent Memory System" similar to the hippocampus. This system integrates not only text, images and speech but also physical signals such as sensor data to enable AI to "remember experiences and habits" like humans.

In view of the problems of Physical AI, Red Bear AI has proposed a systematic solution based on "Memory Science + Full - Modal Large Model":

• Perceptual Reconstruction: Non - invasive and Cost - effective Adaptation

Red Bear AI's full - modal perception system emphasizes the principle of "non - invasiveness". For example, in the intelligent household environment, the system can determine requirements based on modality data such as radar, infrared and speech instead of relying on camera surveillance to maintain the balance between privacy and user experience. By incorporating memory science, AI can remember the specific preferences of users (e.g., the noise sensitivity of the elderly) and achieve "appropriate perception". In addition, Red Bear AI partially replaces expensive high - resolution cameras with cost - effective sensors through the complementarity of multimodal data, which significantly lowers the threshold for the modernization of traditional industries.

• Decision - Making Evolution: From Mechanical Matching to Experience Inference

Red Bear AI's decision - making system has the abilities of "experience replication" and "multimodal agent cooperation inference". The system can store the complete scenarios (input, process, result) of previous decisions and quickly retrieve experiences and optimize decisions in similar scenarios. Taking intelligent veterinary medicine as an example, the agent can summarize behavioral changes, excretion data and the experiences of veterinarians in the memory system to make a three - dimensional decision that is more accurate than one - sided judgment.

• Action Implementation: Closing the Loop and Reconstructing the Industrial Process

Red Bear AI tries to transform the traditional linear model of "R & D - Production - Sales" into a closed - loop of "Data Acquisition - Decision - Implementation - Experience Aggregation". In retail, the "Customer Service Optimization System" can not only process consultations from different channels but also aggregate the experiences of handling difficulties in a knowledge base to continuously improve the response efficiency.

Official Website

Corporate Data Disclosure

It is known that Red Bear AI's Series A+ financing is currently under negotiation and is expected to be completed between July and August. The enterprise value is estimated to be 3 billion yuan. The company has completed several rounds of compliance reviews, including the capital amount, partner agreement, representative arrangements and cost - structure allocation. Red Bear AI has released a series of financial data:

• 2025 Results: The contract value was 250 million yuan, the confirmed revenue was 135 million yuan, the profit margin reached 13% and the profit was about 18 million yuan.
• Gross Margin: According to an audit by Ernst & Young, the gross margin of the pure software business is between 60% and 78%.
• Future Expectations: The sales target for 2026 is a confirmed revenue of 500 - 600 million yuan. It is expected that the pure ARR (subscription model) revenue will exceed 100 million yuan in August or September this year.