Hard Kr Exclusive Premiere | Beihang-affiliated AI new materials team has raised tens of millions of RMB in financing, with its AI Brain + autonomous robotic laboratory already commercialized.
Hard Krypton learned recently that Beijing Heguang Intelligent Technology Co., Ltd. (hereinafter referred to as "Heguang Zhicheng"), an AI new material enterprise, has completed tens of millions of yuan in angel round financing, which will be mainly used for team expansion and independent laboratory capacity expansion.
Heguang Zhicheng was founded in December 2025, committed to making material R&D simpler and faster by introducing AI into the field. Despite its short history, the company has carried out cooperation in new energy, aerospace, pharmaceutical polymer materials, drug modification, optoelectronic materials, photoresist and other fields, and its products have achieved commercialization and completed delivery to multiple customers.
Its chief scientist Liu Yuzhou is a professor and doctoral supervisor at the School of Chemistry of Beihang University. He once served as an R&D scientist at the headquarters of Milliken & Co. in the United States, has published original papers in Science and Nature sub-journals, and holds more than 40 authorized international patents. Chief Scientific Advisor Yang Wantai is an academician of the Chinese Academy of Sciences, who has long been committed to the methodological research on polymer material synthesis and modification chemistry. He is currently a professor in the Department of Chemical Engineering of Beijing University of Chemical Technology and Tsinghua University, and the director of Beijing Laboratory of Biomedical Materials.
At present, the R&D of cutting-edge new materials has entered the deep water zone of complex formulas and extreme synthesis conditions, and the efficiency and effect of the traditional R&D model relying on manual experience and trial and error have reached the ceiling. At the same time, the continuous breakthrough of downstream cutting-edge technologies and the accelerated iteration of the industry have brought an unprecedentedly urgent cycle for the iteration of new materials. The downstream is forcing the 10-trillion-yuan material industry to embrace the transformation wave of AI to accelerate the accurate and efficient R&D of materials.
Different from manufacturers on the market that only provide simulation computing tools or only sell automated equipment, Heguang Zhicheng has realized the closed-loop integration of "computation-experiment-data", and its core capabilities and products correspond to three major sectors respectively.
The first is Yun Yunsuan™, an intelligent agent for material design and synthesis. In the nearly infinite chemical space, the agent can quickly complete massive molecular screening, find the molecular structure that meets user needs and the corresponding effective synthesis path, reduce offline experimental trial and error from the source, and greatly lower the access threshold for AI material R&D.
The other core sector is the AI-driven autonomous laboratory. According to Liu Yuzhou, it can be understood as a fully automatic unmanned synthesis experimental factory, where AI independently completes the collection of experimental data and multi-dimensional parameter adjustment. At present, Heguang Zhicheng has launched two specifications of autonomous laboratories at the milligram level and kilogram level, overcoming industry problems such as high-precision feeding of multi-form solids and automation of anhydrous and oxygen-free reactions, realizing 7×24 hours unattended high-throughput synthesis. Its single-day experimental throughput is equivalent to that of a multi-person R&D team, and it has complete data production and traceability capabilities.
For the material industry, data is the most critical core element, and the high-quality data production capacity is exactly the third core capability sector of Heguang Zhicheng. Data determines the accuracy of the AI model. There is a systematic deviation between calculated data and real wet experimental data. Only by integrating a large amount of real experimental data can the model move from the "armchair strategist" prediction to industrial-level accuracy that can be synthesized and verified. Through the fully automated autonomous laboratory, Heguang Zhicheng is continuously producing high-quality chemical/material data. At present, it has accumulated PB-level chained, multi-modal, full-sample chemical data, which is the key to the algorithm accuracy of Heguang Zhicheng reaching the industrial level.
Through the technical closed loop of computing agent, autonomous laboratory and high-quality data set, Heguang Zhicheng can greatly shorten the material R&D cycle and reduce the material R&D cost. Liu Yuzhou gave an example that when Heguang Zhicheng cooperated with a well-known listed enterprise in the optoelectronic material industry, it optimized the structure of a key material that had been put into large-scale downstream production, and greatly improved the reaction conversion rate without changing its production line and reaction process.
In addition, different from ordinary material R&D service providers, Heguang Zhicheng supports a very wide range of material types, which is related to the "atomic world model" it has built. This model allows AI to understand the underlying microscopic laws such as the interaction force between atoms and chemical bond breaking. "Whether it is polymer, medicine, catalysis, luminescence or photoresist, the synthesis of molecules will eventually fall to the atomic level. These underlying logics and data are universal and transferable. Therefore, we can quickly solve all customers' problems," Liu Yuzhou said.
Liu Yuzhou said that Heguang Zhicheng can be understood as the embodied intelligence in the field of materials and chemistry. "The brain thinks accurately based on a large amount of real, high-quality data to find the best synthesis path and catalyst. The embodied part puts the ideas of the brain into practice through robotic arms, special chemical reaction equipment and testing equipment, continuously optimizes them, produces deliverable real material samples, and synchronously accumulates new data to feed back to AI." At present, the enterprise has an automated laboratory of more than 1,000 square meters in Beijing, which has realized full-process automation of feeding, reaction and detection. It can complete more than 400 full-process experiments and large-scale data collection every day. "Making the embodied intelligence in the material field more quickly popularized in the industry is the key direction that the company will focus on in the next step."
The following is an excerpt from Hard Krypton's interview with Liu Yuzhou, founder of Heguang Zhicheng:
Hard Krypton: You used to be a professor at the School of Chemistry. How did you decide to start this business in the first place?
Liu Yuzhou: I have always been very interested in the industry, and I want to know where the knowledge and abilities I have learned can create value. After graduating with a doctorate in the United States, I joined Milliken, an American material R&D company, mainly engaged in special chemicals and polymer additives.
I originally thought that R&D was to do experiments day after day to get results, but during my time in the United States, I was greatly shocked to find that they use a large amount of data for design to make R&D more purposeful and accurate, so as to maintain a long-term leading position in the global industrial sector. This gave me a huge conceptual impact, and made me start to think about whether data can be used to accelerate material R&D.
After returning to Beihang University, I had a lot of discussions with teachers from the School of Computer Science and the School of Automation. Combined with the development of artificial intelligence's large-scale data processing capability, computing power and experimental technology in the past ten years, high-quality data collection has become possible. Combining my own industrial and academic experience, we started this business, and the verification from customers has confirmed that this new method can play a very obvious accelerating role.
Since 2014, we have maintained very close connection with industrial customers, fully understood their needs, and continuously developed the capabilities of relevant solutions. Therefore, our understanding and connection of customer scenarios started many years ago.
Hard Krypton: The core barrier in the industry lies in data, but in the initial stage, how did you access and collect those unpublished or failed experimental data?
Liu Yuzhou: Data is indeed crucial. The biggest pain point in the material industry is that although there are hundreds of millions of data in public databases, 30% of them may be false, and professionals can hardly distinguish the authenticity, so they cannot be directly used for deep learning training.
After many detours, we found that we must build our own autonomous laboratory. We have established a complex characterization system that can mutually correct each other, avoid generating false data, and ensure high quality and high dimension of data. We collect all data, whether successful or failed. In addition, the calculation results of our dry experiments (computer models) have been strictly compared with real experiments, and both the virtual and real experimental models are very accurate.
It is precisely because we have dry and wet experimental data and are deeply embedded in various actual scenarios of customers that when customers have needs, we provide them with the final solution to the problem, rather than just selling a set of software or a piece of equipment to customers for them to calculate and operate by themselves.
Hard Krypton: You once mentioned that there is a phenomenon of "valuing prediction while neglecting synthesis" in the current industry. What is the real significance and pain point of synthesis?
Liu Yuzhou: The field of materials includes two aspects: one is to predict properties, and the other is to find synthesis schemes.
Prediction seems simple, which can be done by computer simulation, but the Schrödinger equation and quantum mechanics involved cannot be solved accurately, and there are deviations. The predicted results are often difficult to be directly used in real scenarios, and customers cannot see clear value, so it is difficult for them to pay for the products.
However, synthesis is the core pain point that customers care most about. First of all, the stock market is huge. China has a 10-trillion-yuan new material market. Customers hope to make existing products better, lower in cost and faster in synthesis, which requires optimizing the existing process routes. Second, many new ideas in the industry have long existed, but no one can synthesize them. Only by presenting the finished samples can downstream customers recognize them. Therefore, we choose to solve the path of "how to synthesize materials at low cost and industrial scale".
Hard Krypton: What are the company's short-term plans and long-term vision in the next stage?
Liu Yuzhou: In the short term, we hope to use the financing to expand the scale of the platform, improve the computing capabilities of the autonomous laboratory and the intelligent agent, collect more data, and make the model run faster. At the same time, relying on our own laboratory, we will deepen the research of autonomous laboratories, and expand horizontally in the fields that have been implemented to serve more customers.
In the long run, we hope to build the platform into a general-purpose material R&D platform. Just like people can easily program with AI tools now, in the future, ordinary people or small and micro users can spend a small amount of money to purchase tokens, and then automatically complete the design and synthesis of materials on the cloud according to their own needs, so that everyone can create new substances according to their own ideas.