Economic Observer reporter Zhang Ling

In the field of AI pharmaceuticals, providing tools for making drugs is a less traveled path compared to betting on drug pipelines. In the eyes of some industry insiders, it is difficult to support a large company solely by making tools. In the past few years, more and more AI pharmaceutical companies, including Yingsi Intelligent (03696. HK), Yitai Technology (07666. HK), Anew Labs, Huashen Zhiyao and other star AI pharmaceutical companies, have extended their business scope to the drug pipeline. Baitu Shengke (Beijing) Intelligent Technology Co., Ltd. (hereinafter referred to as "Baitu Shengke") is one of the few companies that choose to stay upstream and provide infrastructure for AI pharmaceuticals.

They make medicine and make big money; we provide tools and make relatively small money, but we can serve more customers, "said Liu Wei, co-founder and CEO of Baitu Shengke, in an interview with the Economic Observer in October 2026.

Liu Wei has 18 years of early investment experience, having served as a partner at Lenovo Star and the founding CEO of Baidu Ventures. He has invested in and incubated multiple early-stage technology companies. In 2020, he and Robin Lee, the founder of Baidu, jointly founded Baitu Biotech, which cut into AI Pharmaceuticals from the basic model of life science, took proteins, genes, cells and even organs as research objects, input, predict and obtain values of their various attributes according to their biological sequences, and unified them into one query platform.

Liu Wei said that the third and fourth years of the establishment of Baitu Biotechnology were the most difficult times. At that time, Baitu Biotechnology continued to invest in platform technology research and development, but had not yet produced concept validation results for drug discovery and did not receive much recognition from users, resulting in significant internal and external pressure. He described the process as like trekking through an uninhabited area.

Now, as the underlying AI capabilities of Baitu Biotechnology are validated in multiple projects, it has become a trend for pharmaceutical companies to build their own AI capabilities. Whether Baitu Biotechnology's choice can run smoothly will be subject to commercial testing.

|Dialogue|

Economic Observer: From the perspective of customer demand and business implementation, the development of Baitu Biotechnology seems to be faster this year than in previous years.

Liu Wei:Yes. One reason is that our underlying technology and product packaging have become more mature, and another reason is that the industry's investment in innovative drug research and development is growing rapidly.

In the past year or two, many biotech companies have made money through BD (Pipeline Authorization Cooperation) transactions. Now, they are starting to reinvest these funds into more innovative drug research and development. If they are developing drugs that can be effectively developed using mainstream biotechnology, they may be able to do it themselves or hire CRO (Biological Research and Development Outsourcing) companies; But if they want to enter a completely new field, mainstream biotechnology has a low success rate, and they are more concerned about how to build their own unique research and development platform, so that AI can play a key role in this platform from the beginning.

Economic Observer: Since its establishment, which stage has been the most difficult for Baitu Biotechnology?

Liu Wei:Each stage is actually quite difficult. For example, this year, the industry is very hot, and the difficulty that comes with it is that there are many customer demands, involving the development of many new models. Our underlying technology is relatively mature, and we need to implement, iterate, and expand in many fields, while also addressing organizational management issues.

If we have to say the most difficult times, they are probably the third and fourth years of the company's establishment.

When the company was first established, people would think that we were entering a new field, and financing and recruitment were relatively smooth. But after the actual implementation, the development of underlying technology will go through a cycle of 1-3 years, and many indicators may not improve so quickly, and some indicators may even have a distance from the industry's perception at that time. That process was like trekking through an uninhabited area.

At that time, we had not yet produced molecular concept validation results and had not received much recognition from users, resulting in significant internal and external pressure. People often ask me, why not pursue mainstream small molecule and standard antibody drugs? If we do those things, AI may already be able to achieve them at that point in time.

Economic Observer: You started investing in AI+life sciences 10 years ago, why did you establish Baitu Biotechnology in 2020?

Liu Wei:2020 is the first year of human basic big model technology. In 2018, the GPT-1 released by Open AI had only over 100 million parameters and the effect was not significant. By 2020, GPT-3 had been iterated with 175 billion parameters. There has been a huge change in the underlying technology, which is a good opportunity to become a basic modeling company that can solve hundreds or thousands of problems in one package. Biological modeling is cutting-edge enough that we cannot find investment targets globally, so we have decided to take risks on our own.

Economic Observer: Did you consider not establishing an independent company at that time, but instead doing it as a business of Baidu Group?

Liu Wei:No. This is a job starting from scratch, inherently a startup company. More importantly, in the data of life sciences, a considerable proportion must be produced by oneself.

We have increasingly entered high-throughput laboratories and even the research and development of upstream instruments. Just as autonomous driving and embodied intelligence companies must focus on upstream key components and sensors, we also need to gradually move upstream. It is not realistic to put it inside a big Internet factory. Algorithms, applications and agents can be built in large factories, but it is difficult to develop the underlying equipment of life science. The pace of research and development and the way of investment it requires are different from the iterative logic of Internet companies.

Economic Observer: What specific culture and logic are far apart?

Liu Wei:The underlying equipment and data of life science need long-term investment, and it often takes several years to see the results. It is difficult to follow the iterative method of Internet companies in this R&D rhythm.

Economic Observer: Zhang Yiming, Huang Zheng, Wang Xiaochuan and others have expressed similar views - dreams are life sciences. Why is this?

Liu Wei:Life sciences is an industry with an annual output value of several trillion US dollars, with significant economic and social value, and a strong embrace of innovative technologies.

Why has AI entered this field more frequently recently? Because although this field is large, it was not easy to enter before. AI needs to have the ability to digitize the physical world before it can be put to use. In the past, life sciences were more focused on traditional experiments, with little data accumulation. In recent years, various new omics technologies have emerged that can collect more data, making it possible for everyone to enter.

Economic Observer: What do you think about the recent financing of Anew Labs under ByteDance?

Liu Wei:Firstly, congratulations. After becoming an independent company, it can focus more on its work, measure its main business more clearly, and recruit more personnel. Now, Anew Labs has a clear positioning as a company that uses its own AI technology to build pipelines.

In the field of AI pharmaceuticals, different companies choose different development paths. One type is based on self-developed pipelines and directly participates in drug research and development; Another type focuses on life science big models and platform capabilities, providing underlying AI capabilities for biopharmaceutical companies. The two types of paths correspond to different business logics and success conditions.

For companies with self-developed pipelines as their core, the ultimate success of the company depends on the selection of pipeline projects, targets, and indications, and sometimes also requires luck. If the wrong pipeline is chosen, no matter how good the AI is, it cannot be effective.

In today's industry, whether it's companies with AI backgrounds entering the pipeline field or many mid to large biotech and multinational corporations recruiting their own AI teams, they are essentially doing the same thing: building AI capabilities and serving their own drug pipelines. Anew Labs is more on the same track as these pharmaceutical companies.

We have a clear stratification in the industry chain with them: they make drugs and make big money; We provide tools to earn relatively small amounts of money, but can serve more customers. In a sense, Anew Labs can be our customer, just like many AI pipeline companies. Because their main focus is on making drugs, although they can establish some AI research and development capabilities, it is impossible for them to do everything themselves, and there is no need to do a lot of low-level things all by themselves for the needs of several pipelines.

Economic Observer: Has Baitu Biotechnology considered making medicine in the past?

Liu Wei:No. There are many pipeline companies in the field of life sciences, and they are growing rapidly. Companies that provide tools or CRO services, such as WuXi AppTec (603259. SH/02359. HK) and WuXi Biologics (02269. HK), have a higher market value than many biotech companies. Because platform companies do not share the success of a single enterprise or pipeline, but rather the success of many customers.

Economic Observer: Why did we co establish an AI pipeline research and development company with Nippon Paint this year?

Liu Wei:Platinum Pharmaceuticals embraced AI earlier. We have been working together since 5 years ago. Our early AI technology empowered multiple innovative AI pipelines and achieved good validation results.

Based on our underlying foundational model, it can form its own unique AI model capability that continuously evolves through large-scale pipeline combinations, improving the overall success rate of innovative drug development, rather than relying on probability to build the next pipeline after creating one pipeline.

I think this is a change in mindset. When creating First in class drugs, if the project is done too little, the failure rate is too high; Doing too much, but the cost is too high. Relying solely on traditional experimental methods is difficult to resolve this contradiction. Mass production based on AI native platforms will become an important paradigm.

Economic Observer: Will some traditional pharmaceutical companies easily have doubts about this cooperation model?

Liu Wei:I prefer to refer to them as mainstream pharmaceutical companies rather than traditional pharmaceutical companies. Over the past many years, these enterprises have been able to create excellent pipelines using mainstream biotechnology and existing IT technologies, and have achieved success. Throughout the entire drug development process, there may not be a need to call upon advanced AI models.

In the past two years, the innovative pharmaceutical industry has undergone many changes: pharmaceutical companies have more confidence in AI, and people increasingly feel that AI cannot be missed; The research and development field of pharmaceutical companies is also moving towards emerging directions, and more and more pharmaceutical companies want to be First in class; The AI capabilities of pharmaceutical companies themselves are also maturing. Many pharmaceutical companies already have AI application engineers who can use AI for auxiliary design, and even have application level AI models and AI software development engineers.

(Author Zhang Ling)