Shenwan Hongyuan Group: Tech giants enter AI4S; watch for changes in the upstream industry chain

Shenwan Hongyuan Group: Tech giants enter AI4S; watch for changes in the upstream industry chain Shenwan Hongyuan Group released a research report stating that, according to data, the global AIDD market size increased from US$5.4 billion in 2019 to US$11.9 billion in 2023, and is expected to reach US$74.6 billion by 2032, with a 2023-2032E CAGR of approximately 22.6%. Compared with small molecules, large-molecule AI pipelines are less mature, and a value-depression landscape is emerging. Wet-lab validation has become the rate-limiting step in model development, and related targets are recommended for attention. The main views of Shenwan Hongyuan Group are as follows: Tech giants are entering AI4S through a "computing power + large models + wet-lab closed loop" approach According to Frost & Sullivan data, the global AIDD market size increased from US$5.4 billion in 2019 to US$11.9 billion in 2023, and is expected to reach US$74.6 billion by 2032, with a 2023-2032E CAGR of approximately 22.6%; on the policy side, AI4S was listed first by the State Council among "key actions to accelerate implementation," and the "15th Five-Year Plan for the Development of the Pharmaceutical Industry" further clarified the AI-driven transformation of R&D and production paradigms. Against this backdrop, tech giants are accelerating their entry. Among tech giants, Anthropic launched Claude for Life Sciences and built its own wet lab; Isomorphic Labs built IsoDDE based on AlphaFold3 and has reached cooperation with Eli Lilly, Novartis, and Johnson & Johnson; domestically, BioMap's xTrimo V4, Xinsheng Laboratory's AnewFold, and others are also catching up rapidly. Tech giants have chosen large-molecule models as the breakthrough point for multiple reasons First, the "underlying infrastructure" of small-molecule AI is already relatively mature, and the "model layer" and "data layer" for small-molecule generation and structure prediction are no longer scarce resources. Second, small-molecule patent barriers are strong and difficult to circumvent, while large molecules such as antibodies still have considerable room for circumvention through differentiated CDR sequences and epitopes. Finally, compared with small molecules, large-molecule AI pipelines are less mature, and a value-depression landscape is emerging. Currently, AI drug clinical pipelines remain dominated by small molecules, with large-molecule pipelines accounting for only 28%. Moreover, large-molecule AI clinical pipelines are mainly concentrated in Phase I/II (82%), with only Generate's TSLP monoclonal antibody GB-0895 having entered Phase III. However, this antibody is not a completely new antibody scaffold designed de novo by AI, but rather one that has been engineered and optimized by AI. Wet-lab validation has become the rate-limiting step in the development of large-molecule AI models, and upstream "picks-and-shovels" players will benefit directly Large-molecule AI models face structural difficulties such as large molecular weight, long sequences, heterogeneous data, and slow DBTL cycles, and wet-lab validation has become the rate-limiting step in model development. After large-scale candidate generation becomes routine, conversion efficiency will depend on the ceiling of wet-lab throughput, and upstream "picks-and-shovels" players will benefit directly. Among them, Twist guides AI-related orders to achieve triple-digit growth in FY2026; GenScript's AIDD-related business doubled year-on-year in 2026H1. The bank recommends paying attention to related targets, including Twist, GenScript Biotech, Acrobiosystems, Novoprotein Scientific Inc., and others. Risk warnings: risk of wet-lab validation demand falling short of expectations, risk of technological iteration falling short of expectations, risk of model commercialization falling short of expectations