Citi: AI drug discovery accelerates candidate generation; Citi sees opportunities in lab validation and drug assets

AI is generating more drug candidates. Why are laboratories becoming more important?

Citi surveyed 50 US pharmaceutical R&D technology decision-makers. Of those, 98% were at least testing AI, and 72% had expanded deployment or completed deployment at scale. As AI generates candidate molecules faster and in greater numbers, the laboratory capacity needed for synthesis, protein expression and testing could become a bottleneck.
Citi expects lower marginal molecular design costs to allow drugmakers to explore targets and hypotheses that were previously uneconomical, expanding the pool of R&D projects. Experimental results can then be fed back into models, driving another round of design and testing.
Outsourced experiments are not the only way to benefit: AI drug discovery companies may bring some early experiments in-house to speed up model feedback. By contrast, downstream biologics development and manufacturing require more upfront capital, process scale-up capabilities and quality systems. Citi believes the economic incentive to outsource these activities to established service providers is stronger.

Citi expects three models to coexist: drugmakers' own networks use AI without transferring data across organizations; technology workbenches sell models, computing power and R&D tools; and integrated drug generation platforms advance candidates and benefit through licensing, milestone payments and retained asset interests.
Citi believes those with differentiated data, high-throughput experimental validation capabilities or drug assets are better positioned to retain a larger share of customer spending. Foundation model companies are also building laboratory capabilities: Anthropic has established an internal wet lab, but has not separately disclosed revenue from life sciences products or drug R&D output.

Citi is positive on WuXi Biologics' prospects for protein expression, screening and subsequent development and manufacturing work. The company's protein generation services grew by more than 40% year over year in the first half of 2026; management estimates full-year revenue for the business at less than 200 million US dollars. Citi calculates that this is equivalent to about 10% of preclinical revenue in 2026.
Protein generation services have a gross margin of 50%–60%, above the company's overall margin of 45%–50%. Management also says average selling prices for bispecific antibody and ADC projects are 50%–100% higher than for traditional products; Citi believes a higher share of complex projects supports gross margins.
In the first half of 2026, the company added 169 integrated projects, of which 123 were organic additions, up 43% year over year, and another 46 came from acquisitions. Of newly signed projects, 71% were complex types such as bispecifics, multispecifics or ADCs. Management says preclinical project prices rose 10% in the first half and, given delivery cycles of 6–9 months, the impact of pricing will be more visible in 2027 revenue.

Citi sees value in Insilico Medicine's platform and asset partnerships. Since 2021, the company has nominated a cumulative 34 preclinical candidates and has a pipeline of more than 40 programs and 13 approved investigational new drug applications. Its commercialization approaches include out-licensing, co-development and platform discovery partnerships.
The partnership with Takeda includes approximately 60 million US dollars in project initiation fees, near-term payments and milestone payments, with a total deal value of up to approximately 600 million US dollars, plus a share of sales. Citi believes the multiple clinical-stage and filing-stage programs provide opportunities for recurring pipeline partnerships.
The next step for the platform is customer monetization. Its MMAI Gym contains more than 1000 drug R&D benchmarks. A model developed with Liquid AI can be deployed on drugmakers' private infrastructure, allowing customers to keep proprietary molecular, experimental and target data on site. Citi expects the business model to combine private deployment, specialized fine-tuning, evaluation services and joint model development.