Cathay Pacific Haitong: The scientific model enters the core R&D process of pharmaceutical companies, and the demand for R&D outsourcing is expected to continue to rise

The Zhitong Finance App learned that Cathay Pacific Haitong released a research report saying that on the big model side, the big scientific model has entered the core R&D process of pharmaceutical companies. The bank believes that data generation and verification capabilities with the “wet and dry loop” as the core are becoming a core competitive barrier for AI pharmaceuticals, and that wet test production capacity and high-quality data have become scarce strategic assets; under the resonance of large models, pharmaceutical companies, and service providers, industry prosperity and R&D outsourcing demand are expected to continue to rise.

Cathay Pacific Haitong's main views are as follows:

Big model side, big scientific model enters the core R&D process of pharmaceutical companies

The overseas model was upgraded from a “tool” to a “smart body+wet experiment”: Anthropic established a wet laboratory in the San Francisco Bay Area and worked in parallel with about 950 Claude agents for 21 hours to independently discover a previously uncharacterized bacteriophage enzyme system (ART); OpenAI led biosafety company RedQueen Bio into AI biological defense; Isomorphic Labs' iSodde engine completed functional molecular design in 2-4 days in the form of an intelligent body. While iterating model capabilities, domestic manufacturers are building unique data moats in the segmented field: Insili Intelligence announced the full deployment of “smart layers” and MCP, Jingtai Technology upgraded the XtalPi Science platform, and Huada Intelligent Manufacturing and Baitu Biotech jointly built a large cell model.

On the pharmaceutical enterprise side, MNC uses “self-building+cooperation” to accelerate AIDD capacity building, and ecological cooperation is being implemented intensively

The overseas layout can be classified into three categories: the model enters the main R&D process, the joint construction of a wet test network ecosystem, and cooperative professional models to accelerate development in specific fields. For example, Novartis introduced Claude Science into the R&D process, Lilly collaborated with Twist, Ginkgo, and Kingsley to build a closed loop of TuneLab data. Roche announced “AI is autonomous and controllable” and disclosed that 40% of pipeline decisions already have traceable AI/computing contributions. The list of domestic transactions continues to grow: the potential total value of the nine BD cooperation contracts disclosed by Insili Smart during the year was about US$7.3 billion, the potential milestone of cooperation between Jingtai Technology and DoveTree reached a maximum of US$5.89 billion, and the small oral GLP-1 molecule designed by Derui Zhi Pharmaceutical AI reached a commercial cooperation of up to 1.75 billion yuan.

On the wet test side, construction is being upgraded from “outsourcing services” to “strategic infrastructure”, and the “arms race” for global production capacity is accelerating

According to the bank's judgment, the faster the AI closed loop changes, the stronger the dependence on wet test throughput, and the growth rate of laboratory demand is expected to be faster than the number of AI drug projects. Overseas, Roche built an independent AI laboratory, Anthropic built a biological laboratory, and Eli Lilly and Nvidia jointly built a joint innovation laboratory with a maximum of 1 billion US dollars over five years, and Recursion maintained a weekly wet test throughput of about 2.2 million times a week and 65 pB of proprietary data; domestically, Kingsley distributed a net capital raised about HK$2.33 billion, and about 70% invested in AIDD wet test production capacity. Jingtai Technology operated tens of thousands of square meters of automated laboratory clusters. Baitu Biotech and Platinum Pharmaceuticals jointly built a closed loop dry and wet laboratory system. Insili announced the LabClaw intelligent laboratory system for production It can be expanded and increased simultaneously with data infrastructure.

Risk warning: model iteration and clinical progress fall short of expectations, implementation of cooperation and commercialization fall short of expectations, increased industry competition, changes in data compliance and regulatory policies