Key Highlights

  • Drugmakers say AI has cut by as much as half the time to produce new molecules and reach the clinic.
  • Sanofi said AI saved at least a year developing asthma treatment lunsekimig by letting it skip a dose-selection study.
  • Novo Nordisk said AI cut design cycles from about eight to three and shortened each from eight weeks to four.
  • Researchers counted 129 AI-enabled drug candidates across 69 companies in March, up from around 117 eight months earlier.

AI Moves Beyond Paperwork

Artificial intelligence has sped up early-stage Drug Development at some of the world's largest pharmaceutical companies this year, cutting by as much as half the time it takes to produce new molecules and move them into the clinic, according to their top scientists. GSK, Bristol Myers Squibb, Novartis, Eli Lilly, Novo Nordisk, AstraZeneca and Sanofi say they have moved beyond less glamorous AI tasks such as preparing regulatory filings. They now use the technology to design molecules, sort the most promising ones and identify safety risks.

An AstraZeneca executive said the company uses AI in every part of the pipeline from discovery through development.

Concrete Results Cited

Sanofi said it saved at least a year developing the asthma treatment lunsekimig using an AI-enabled approach that translated early biology into clinical decisions and let the company skip a dose-selection study. Novartis said AI tools helped it find five targets it prioritized for kidney disease treatments. Its head of biomedical research said the average time from project start to a drug candidate has fallen from three years to about 18 months to two years.

Bristol Myers said one of its sickle cell anemia candidates owes its development to its AI systems. Its chief research officer said the company uses AI to refine molecules and run simulations on trial data to decide which molecules progress, and that the next area is tying those models to laboratory automation. Novo Nordisk said AI cut the number of design, Manufacturing and testing cycles for new molecules from around eight to three and shortened each cycle from eight weeks to four.

Limits on the Evidence

Clinical trials and experiments still take time. Historically, nearly 9 out of 10 drugs developed using conventional methods Fail in clinical trials, and bringing a drug to market can cost billions of dollars and take 10 to 15 years, according to the industry lobby group PhRMA. The number of AI-enabled drugs in clinical testing has risen only incrementally: researchers identified 129 candidates across 69 companies in March, up from around 117 eight months earlier. Most programs remain at an early stage.

An oncology fellow at the University of Chicago who helped compile the data said he sees a sharp rise in AI drugs in preclinical development and trials over one to three years, but that he does not know whether those drugs are more successful than non-AI drugs, and he does not think they are approved any quicker, because of bottlenecks with human trials and regulators.

Anthropic as Partner, Vendor or Rival

Anthropic, which listed on the Nasdaq this week, has opened its own laboratory and made investments aimed at speeding drug discovery. It works with partners including Bristol Myers, launched science-focused software called Claude Science and added Novartis chief executive Vas Narasimhan to its board. Novo Nordisk and other companies use Anthropic models in their Research and Development work.

At the Nasdaq on Thursday, Anthropic chief executive Dario Amodei spoke of his father dying of a disease shortly before a cure was found. "I believe these processes can be substantially accelerated by AI," he said. A leader at Lilly's TuneLab program said claims that AI can cure all disease overstate current capabilities, and that curing even one disease is the place to start before talking about the rest. She said Lilly once saw such companies as vendors and increasingly sees them as collaborators. GSK's chief scientific officer said he regards Anthropic as one more competitor.

Data as the Dividing Line

Novo Nordisk and Sanofi said they do not provide proprietary data to frontier labs to train their models. A Sanofi research executive said that without Training data a model degrades very fast and that he expects AI to cut early-discovery timelines by 20% to 25% within two years and perhaps by half within five years across more than 100 programs. A Venture Capital investor said Anthropic and rival labs are shopping for pharma-related acquisitions. Anthropic did not respond to a request for comment on that point.