The Zhitong Finance App learned that Cathay Pacific Haitong released a research report saying that the difference between this round of AIDD and the previous round is that the AI upgrade from search efficiency tools to being able to implement a complete decision-making system and design elements from scratch is expected to change the underlying logic of drug design. Since 2026, basic model manufacturers have increased their investment in the medical industry. The pharmaceutical enterprise cooperation model has changed from a project system to in-depth enterprise-level co-construction such as model subscriptions and joint laboratories, and the industry has entered a key window for clinical verification and order fulfillment.
Cathay Pacific Haitong's main views are as follows:
AI pharmaceuticals ushered in an inflection point in the industry, and big models entered the market to restructure the industry paradigm.
The rapid development of big language models has accelerated scientific research discoveries, including in the field of biomedicine. The bank believes that the difference between this round of AIDD and the previous round is that the AI upgrade from search efficiency tools to being able to execute a complete decision-making system and design molecules from scratch is expected to change the underlying logic of drug design. Since 2026, basic model manufacturers have increased their investment in the medical industry. The pharmaceutical enterprise cooperation model has changed from a project system to in-depth enterprise-level co-construction such as model subscriptions and joint laboratories, and the industry has entered a key window for clinical verification and order fulfillment.
AIDD is expected to improve the pharmacogenicity of candidate molecules and expand the space for biological hypotheses.
The core pain point of pharmaceutical research and development is clinical transformation. The success rate of corrected drug-indication programs has remained low at about 8%-9% in recent years, and about 86% of clinical sinking costs come from projects terminated in phase II and later. The bank believes that most of the lack of efficacy and safety issues are rooted in early decisions on targets and candidate molecules. AIDD is expected to break the game from both ends. On the one hand, it simultaneously optimizes various molecular properties and enhances pharmacogenicity; on the other hand, it broadens the selection space for targets, epitopes, and molecular forms, and expands the number of projects entering the R&D funnel. As a result, there is a double increase in market space. First, the increase in clinical success rate is expected to restructure the R&D cost model, and the saved R&D expenses can be invested in more projects; second, the number of projects themselves can be expanded.
A closed loop of computation and experiments is formed, and wet testing services benefit definitively.
The bank believes that the development of small molecules on existing drug targets is relatively adequate, and that new projects rely more on breakthroughs at the target and mechanism level; AI design from scratch allows large molecules to continue adding new projects based on the expansion of epitopes and molecular forms on known targets, and the experimental label data produced by these projects (including successful and failed samples) is currently the core material for large-scale model training and iterative optimization. As a result, the demand bottleneck will gradually be transmitted downward along the chain of “AI model generation design → DNA synthesis → protein production → experimental verification → pre-clinical testing”.
Orders for AI wet experiments are gradually being fulfilled, and capacity barriers continue to deepen.
Wet experimental data with real labels is currently the most definitive requirement. DNA synthesis and protein production have become the most direct bottlenecks, and Kingsley (integrated delivery of gene-to-protein) and Twist (large-scale sequence synthesis) are the first to benefit. The bank believes that during the model training period, supplier competition focuses on delivery throughput, turnover speed and data quality, and that automation equipment has become the main limiting factor for current production expansion; the focus of medium- to long-term competition shifts to the preparation and characterization of complex targets and innovative modes, and the ability to verify and cover the entire chain. Furthermore, the current pharmaceutical process has not been completely restructured, and demand from all links in the industrial chain will be released in sequence, and back-end links such as efficacy, safety evaluation, and CMC are expected to take on incremental increases as the pipeline progresses.
Risk warning: the risk of technology iteration falling short of expectations, the risk of orders and production capacity falling short of expectations, the risk of increased industry competition, and the risk of data and supply chain security.