Within the recent innovative drug sector, the AI drug discovery niche has quietly broken away from the broader market trend.

Hong Kong-listed Genscript Biotech Corporation has seen steadily rising capital recognition after officially announcing a wet-lab collaboration with Eli Lilly's TuneLab platform. On the A-share side, Nearshore Protein has also shown strong momentum 鈥?though investors should note that the company cautioned in its share price abnormality announcement that AI-related wet-lab business currently accounts for only a small share of revenue, with the stock price reflecting expectations ahead of actual results.

Industry-level signals are more convincing. The Ministry of Industry and Information Technology and nine other departments have just issued the pharmaceutical industry "15th Five-Year Plan," which directly elevates AI-empowered drug development to a national-level strategy. Overseas, Anthropic has reached a deep collaboration with Novo Nordisk, and ByteDance-spun-off AI drug discovery company Anew Labs completed a $290 million first-round financing at a post-money valuation of $1.5 billion. Even more tangible are BD deals 鈥?since the start of this year, global pharmaceutical companies and AI platforms have signed deal after deal, with upfront payments and milestone payments representing real money.

Many people's impression of AI drug discovery remains stuck at "using large models to generate molecules, more gimmick than substance." But by 2026, the industry has quietly passed the proof-of-concept stage, and the underlying picture this time is quite different from the pure concept-driven rally two years ago.

From storytelling to real delivery: the industry has crossed the inflection point

This round of AI drug discovery enthusiasm is not driven by a single event but is the result of a triple resonance of policy, commercialization, and clinical validation.

The "15th Five-Year Plan" for pharmaceutical industry development explicitly identifies AI-empowered drug research, clinical studies, and pharmaceutical manufacturing as key tasks. This means AI drug discovery is no longer a spontaneous corporate exploration but has been incorporated into top-level industrial design, with supporting policies and research investment to follow.

At the same time, major pharmaceutical companies are beginning to pay real money for AI technology, rather than engaging in token framework collaborations. Since 2025, dozens of AI drug discovery-related BD deals have been completed globally, including three platform-level collaborations 鈥?CSPC 脳 AstraZeneca (up to $5.33 billion total), XtalPi 脳 DoveTree (up to $5.99 billion total), and Insilico Medicine 脳 Eli Lilly ($2.75 billion) 鈥?with combined amounts exceeding $10 billion. Leading pharmaceutical companies such as Sanofi, Merck, and AstraZeneca are all intensively forming partnerships with AI platforms, typically under a "upfront payment + milestone + sales royalty" model, which represents the most direct commercial recognition of AI technology's value.

Breakthroughs on the clinical side are even more persuasive. Insilico Medicine's core pipeline Rentosertib, targeting idiopathic pulmonary fibrosis, completed its first global patient dosing in Phase III clinical trials in September 2026, with plans to enroll approximately 320 patients and topline data expected in 2029. Phase IIa data showed that patients in the 60mg dose group experienced a 98.4ml improvement in forced vital capacity (FVC) from baseline at 12 weeks, while the placebo group declined by 20.3ml, with an inter-group difference of 118.7ml. Compared to the clinical reality that marketed drugs nintedanib and pirfenidone can only slow lung function decline, this AI-designed candidate drug demonstrates potentially differentiated efficacy advantages. This is not paper simulation data but results verified through human trials, marking AI drug discovery's transition from "designing molecules" to "validating drug efficacy."

The real barrier is not just computing power 鈥?it is also the dry-wet closed loop

When it comes to AI drug discovery, many people's first reaction is still about competing on large models and computing power. This is actually the biggest misunderstanding of the industry.

AI in drug research and development is essentially a tool for improving efficiency, not a replacement for R&D principles. The underlying logic of traditional drug development 鈥?"ten years and one billion dollars" 鈥?has not changed. AI's role is to compress target discovery and molecular screening from years to months, front-load failures as much as possible, and reduce late-stage trial-and-error costs. But all AI prediction results must ultimately be verified through wet-lab experiments, and the validation data then feeds back into the model, forming a "design鈥揺xperiment鈥搃terate" closed loop. In other words, no matter how fast dry experiments run, wet experiments must be solidified step by step. AI drug discovery detached from wet experiments is a castle in the air.

This also determines three business models in the industry. 1. Platform service model: providing AI drug discovery tools and technical services to pharmaceutical companies, charging by project or subscription, with the most stable cash flow. 2. Pipeline licensing model: using AI platforms to discover candidate drugs, advancing them to preclinical or clinical stages, and licensing them out to earn upfront payments, milestone payments, and subsequent sales royalties 鈥?this is currently the industry mainstream and the model with the greatest performance elasticity. 3. Self-developed product commercialization: bringing drugs to market oneself to earn sales revenue 鈥?the largest potential space but also the longest cycle, and currently no truly AI-originated new drug has been approved globally.

Furthermore, this means the core barrier of AI drug discovery is not just model parameters but also the dry-wet closed-loop capability. The first layer is the data barrier. Public biological data is available to everyone, but high-quality, experimentally validated proprietary data is key to model performance. Leading companies accumulate massive proprietary data through automated laboratories 鈥?the more data, the more accurate the model; the more accurate the model, the higher the experimental success rate, forming a positive flywheel. The second layer is closed-loop capability. Having only AI models without experimental capability means one can only provide upstream services; having only experiments without AI means efficiency cannot improve. Only by integrating AI design with automated wet experiments to form a data closed loop can R&D efficiency truly be elevated.

The value transmission along the entire industry chain also follows a clear sequence. AI first improves dry experiment efficiency, generating massive candidate molecules, which first drives upstream demand for experimental consumables, reagents, gene synthesis, and model animals, then transmits to preclinical CROs, and finally to downstream pipeline value realization. Therefore, the release of industry prosperity travels progressively from upstream to midstream and downstream.

Investment perspective: finding certainty along the industry chain

At the investment level, AI drug discovery is not a broad rally 鈥?different segments differ greatly in certainty, elasticity, and realization pace.

The top priority is the upstream "shovel seller" segment. The logic is simple: AI exponentially increases the number of trial-and-error attempts, and each attempt consumes consumables and reagents. The number of AI-generated candidate molecules grows exponentially, and corresponding wet-lab demand will surge in tandem 鈥?regardless of whether the final drug succeeds, these experiments must be done and the money must be spent. Leading companies in gene synthesis, recombinant protein reagents, cell culture consumables, model animals, and preclinical safety evaluation will be the first to feel order growth. This segment is like selling shovels during a gold rush 鈥?you don't bet on whose mine will produce gold; you profit as long as more people are mining.

Then there are midstream AI platform companies, which have high elasticity but also high divergence. Two hard indicators are key to tracking: first, the number, amount, and upfront payment ratio of BD collaborations 鈥?the higher the upfront payment, the greater the major pharmaceutical company's recognition of the technology; second, clinical progress of self-developed pipelines 鈥?the later the stage, the higher the value realization. But be careful to distinguish 鈥?many companies merely wrap a traditional business in an AI concept without real BD orders or clinical pipeline support, and their rallies won't last.

There are also some niche segments with expectation gaps, such as AI + drug delivery and AI + solid formulations. Market attention is lower than target discovery platforms, but technical barriers are not low and application scenarios are broad, making them easily undervalued directions.

Risks also cannot be ignored. The biggest uncertainty is clinical falsification: Rentosertib's Phase III topline data won't be read out until 2029, and any setback in between could crush sector valuations. Second is valuation overextension 鈥?on October 7, Hong Kong-listed AI drug discovery targets generally pulled back more than 10%, indicating sentiment remains elevated and the margin for error in chasing highs is limited. Additionally, pipeline progress falling short of expectations, technology route iteration, and slowing BD collaboration pace will all affect the speed of realization.

Conclusion

AI drug discovery is an industry where timelines are stretched: targets take years to validate, clinical trials take years to advance, and commercialization is measured in decades. Current order landing and clinical progress are real, but explosive realization across the entire industry still requires time. Gelonghui Research Institute continues to track AI drug discovery industry dynamics, following BD collaboration landing pace, pipeline clinical progress, and upstream consumable order changes, and breaking down industry chain sub-segments and key company fundamentals. If you are also tracking this wave of AI drug discovery, welcome to follow us and together capture the structural opportunities in this industrial paradigm shift. Risk disclosure: Investing involves risk, and decisions should be made cautiously. The above analysis is for industry research and market strategy reference only and does not constitute specific investment advice. Disclaimer: The copyright of this article belongs to the research team of Beijing Gelonghui Investment Advisory Co., Ltd. (Duan Yuehan: A0160625050003). This report is produced based on principles of independence, objectivity, fairness, and prudence, with information sourced from public materials and collected, summarized, and edited in a lawful, reasonable, and appropriate manner. The stock market carries risks, and caution is required when entering. Any investment advice in this article should not be used as the basis for your investment decisions. You must make independent investment decisions and bear your own risks. Any entity or individual that publishes, copies, disseminates, or reproduces this content without the company's permission will be deemed to have infringed upon rights, and the company will pursue legal liability in accordance with the law.