2026 Zhongguancun Forum Annual Conference | Dialogue with Dina Technology Chairman Chi Haipeng: "AI + Black Light Laboratory" is not a game for the wealthy, and "robot scientists" can help more small and medium-sized enterprises with R&D

（Source: Beijing Business Today）

“Dark lab” typically refers to laboratories where experiments are carried out by robots and automation equipment, enabling unmanned operations and around-the-clock efficient performance. It is suitable for fields that require high-throughput experiments, such as biopharmaceuticals and materials science. By participating in optimizing product formulations and developing new materials, among other efforts, it can significantly improve a company’s R&D efficiency.

With the development of artificial intelligence technology, in recent years, dark labs have begun to be deeply integrated with AI. On May 30, 2025, Beijing issued the “Action Plan for the Innovation and Development of High-End Scientific Instruments,” which clearly called for advancing “AI+” to empower instrument innovation, and made specific plans for the construction of dark labs in particular.

This policy direction has put Beijing Dena Experimental Technology Co., Ltd. (hereinafter referred to as “Dena Technology”), which focuses on “AI+ dark labs,” on the fast track. On March 25, Dena Technology’s founder and chairman, Chi Haipeng, said in an interview with Beijing Business Today’s reporter that, from the company’s early days of entrepreneurship to now, it has shifted from designing and building laboratories to R&D of smart labs and prefabricated labs, and then continued to iterate into an “AI+ dark lab.”

At present, Dena Technology’s “AI+ dark lab” solutions are applicable to R&D and testing scenarios with high repeatability and high levels of precision, as well as to experimental environments with high risk and stringent environmental requirements. They can be widely used in petrochemical companies, new energy enterprises, the pharmaceutical industry, and the food industry.

“Previously, the main driver of productivity in a laboratory was people; in the future, it will become AI and robots,” Chi Haipeng said. In his view, an AI+ dark lab not only boosts R&D efficiency to hundreds of times that of humans, but can also lower the innovation threshold for small and mid-sized enterprises—when experimental costs decrease and R&D cycles shorten, more innovation players will be able to participate in frontier exploration.

This March, Dena Technology was selected as a key培育 enterprise in Beijing’s Action Plan for Future Industry Development, as the leading company being重点培育 under the “Construction of Intelligent Materials Laboratories” project. This shift in status signals that Dena Technology has officially entered Beijing’s core group of future industry players.

The following are excerpts from the interview between Beijing Business Today and Chi Haipeng:

Q: Compared with traditional lab design and construction in the past, what changes has the R&D and implementation of dark labs brought? What implementation cases does Dena Technology’s AI+ dark lab have now?

A: An AI+ dark lab is like an end-point highly intelligent hardware device. Through automation and AI-enabled instrument equipment of various forms, it enables unmanned operation and closed-loop control across the entire process. I think the biggest change is that the core driver of productivity has changed. Previously, the productivity driver in a lab was people; in the future, it will become AI and robots. People need bright lighting, working space, and also need to consider safety issues such as evacuation radii. Dark labs do not require these conditions. They can adapt to dark environments and compact workspaces; they control lighting and temperature and humidity on their own, and even operate in oxygen-free and non-corrosive environments. Compared with the past, AI+ dark labs can effectively improve the quality and efficiency of experiments. We have calculated that its maximum efficiency can reach 480 times that of humans.

Currently, our dark labs have commercialized cases in fields such as energy, materials, and food, running 24 hours a day. We do R&D on new materials and complex liquid formulations, for example, optimizing the formulation for Nestlé’s coffee. In addition, our testing scenarios are also providing 24-hour services for many Fortune 500 companies, such as testing extremely toxic chemicals.

Q: You mentioned that compared with human-led experiments, dark labs generate high-quality data. How does Dena Technology’s AI+ dark lab ensure the accuracy of the data it collects, including how to feed the data back to the AI?

A: Some overseas research institutions have mentioned that the fewer people involved in an experiment, the higher the quality of the generated data. On the one hand, the less people participate, the less subjective and intentional interference there is, and the data is more real. On the other hand, people have high trial-and-error costs when doing experiments, while AI experiments are planned. In the early stage, we discussed with scientists and built upon existing pathways to establish preset decision-making mechanisms, formulation optimization pathways, and so on. Later, the dark lab can run on its own, ensuring that each experiment is not repeated and can help optimize the algorithms. Also, the accuracy and fairness of data within a dark lab are ensured. Once data is generated, no one can modify it. We can encrypt it and even make it into a blockchain.

In this process, all localized standard data produced by the dark lab can be further learned by AI, helping algorithms evolve quickly. Some algorithms can generate new algorithms after one or two experiments; others may need to accumulate data from 40 to 50 groups of experiments to optimize once. The time for technological iteration can be minutes or even dozens of days, depending on the length of the entire experimental cycle. For companies, investing in a dark lab allows them to accumulate a large amount of high-quality data “wealth” for themselves.

Q: At present, does the AI+ dark lab more often get used by enterprises with strong financial capacity? Can small and mid-sized enterprises benefit from it?

A: At this stage, AI+ dark labs can already replace a large amount of basic R&D work. Compared with bigger enterprises with more financial resources, it is more meaningful for small and mid-sized enterprises because it has the characteristics of high efficiency, low cost, and high certainty. When I went to a trade show in Dubai, many companies from less developed countries showed interest in our AI+ dark lab. They lack money and even more lack people. Our lab is like a group of scientists working 24 hours a day; for them, it is a very cost-reducing choice for R&D. So I think this technology is actually especially friendly to small and mid-sized enterprises.

In the future, we also consider building an R&D platform for new materials or chemicals, where scientists can commission our platform to conduct rapid R&D, without needing to hire people, buy equipment, or build a lab—at a very low cost—and the data is real, without any subjective interference by humans. If, in the future, data accumulation keeps increasing, it may even become possible to predict whether something is feasible with AI without conducting experiments, achieving even higher efficiency.

Q: What are the specific business priorities for Dena Technology next?

A: We will build end-to-end, full-ecosystem dark labs. First, it has an AI “brain” that directs multi-modal embodied intelligent automation and instrument equipment to carry out experiments. The original instrument equipment was designed for people; in the future, it will be for AI and automation. That means we need to endow them with embodied intelligence characteristics—such as equipping them with visual recognition so they can make judgments about samples on their own. Also, the direction we are working on is to combine instrument equipment with AI and robots, so it can continuously learn and evolve.

Q: You mentioned earlier that China’s localization rate for high-end scientific research instruments is relatively low. In the context of high-end instruments being “bottlenecked,” do you think combining with AI could become an opportunity for China’s instruments to achieve a leapfrogging “change of route”?

A: It will definitely. As a whole product, a dark lab is judged by the user based on the final result, and the instrument equipment is only one component. We can put domestically produced instruments in as components and add algorithms to the instruments to make them smarter, thereby realizing domestic substitution.

We have now built many AI-assisted systems for atlas recognition, tuning, and repair, to help chromatograph instruments tune curve atlases and perform repairs. Some atlases that are not accurate enough or not detailed enough can be made more complete after adjustments, and may even reach or exceed the level of international comparable instruments.

Beijing Business Today reporter: Zhang Xuwang; Internship reporter: Mao Siy i

A huge amount of information and precise interpretation—exclusively in the Sina Finance app