Artificial intelligence could halve the time spent on early drug development at some major pharmaceutical companies. Molecule design and the selection of drug candidates are accelerating, but human clinical trials and regulatory reviews can still limit how quickly new treatments reach patients.
GSK, Bristol Myers Squibb, Novartis, Eli Lilly, Novo Nordisk, AstraZeneca and Sanofi are using the technology for more than administration or the preparation of regulatory applications. AI also helps them design molecules, select promising candidates and identify safety risks early. Sharon Barr, an AstraZeneca executive, says AI is now used at every stage from discovery through development.
Shorter research cycles
Sanofi saved at least a year in the development of its asthma drug lunsekimig by using algorithms to avoid a dose-selection study. Novartis used the technology to identify five new drug targets related to kidney disease, while Bristol Myers Squibb credits its internal systems with discovering a candidate for sickle cell disease.
At Novo Nordisk, the number of design, manufacturing and testing cycles for new molecules has fallen from eight to three, while each cycle now takes four weeks rather than eight. At Novartis, the average time from project launch to a specific drug candidate has dropped from three years to between one and a half and two years.
The number of AI-developed candidates reaching clinical testing is nevertheless growing slowly. Data from March this year recorded 129 such candidates across 69 companies, up from 117 eight months earlier. Nearly 90% of drugs developed using conventional methods fail in clinical trials, and bringing a drug to market typically takes 10–15 years and costs billions of dollars. Faster preclinical work therefore does not necessarily mean faster final approval.
Partners and competitors
Technology companies are seeking a bigger role in drug research. Anthropic recently opened its own laboratory, launched Claude Science software for scientific research and appointed Novartis chief executive Vas Narasimhan to its board. The source reports that the company went public this week; Dario Amodei, Anthropic's chief executive, said before Thursday's opening of Nasdaq trading that the technology could materially accelerate the discovery of treatments. Venture capital investors say leading AI developers are also looking for pharmaceutical companies to acquire in order to gain access to research data.
Bristol Myers Squibb and Novo Nordisk regard Anthropic as a partner and use Claude Science in research and development. GSK, by contrast, sees the new entrants as direct competitors. Eli Lilly specialists consider claims by AI company leaders that the technology will be able to cure every disease exaggerated.
Drugmakers say decades of research experience and proprietary data provide an important competitive advantage. Neither Novo Nordisk nor Sanofi shares its protected data with outside technology companies for training large language models. Matt Truppo, Sanofi's global head of computational and AI strategy for research and development, says external algorithms quickly become unusable without suitable data. He expects Sanofi's systems, built on internal data, to halve early research time in more than 100 development programmes over the next five years.



