In a sterile Bristol Myers Squibb lab about an hour north of Boston, scientists in scrubs and hairnets transfer living cells into a 2,000-liter stainless steel bioreactor, where they grow them for weeks. The goal is to produce proteins that are genetically engineered to attack disease-causing cells.
Small changes in temperature, light or pH levels can stop cell growth and cause drug shortages that put patients at risk. Scientists would normally have to wait to see what went wrong during this delicate process, but now artificial intelligence is being used to closely monitor important variables – such as temperature and oxygen levels – and alert technicians if there are problems.
Every year the World Economic Forum and McKinsey recognize manufacturers who are at the cutting edge of technology, including artificial intelligence. This year, the Bristol Myers Squibb plant in Devens, Massachusetts, was the only manufacturer in the United States to make the 23rd list.
While U.S. companies typically lead in AI research and capital investment, U.S. manufacturers often struggle to translate these discoveries into increased productivity in factories.
Of the 223 factories listed by the World Economic Forum’s Global Lighthouse Network as of 2018, 14 are in the United States, while 99 are in China. Of the American ones, four are in the pharmaceutical and biological sectors.
“China is growing faster,” said Rahul Shahani, a partner at McKinsey who is working with the World Economic Forum on the initiative. He added: “They have technologists in the factories – hundreds of them – while in the US we are competing for the same talent with Silicon Valley.”
Major U.S. pharmaceutical companies are a rare bright spot in the use of AI Many drugmakers, including Pfizer and Eli Lilly, are investing billions in AI and related technologies to speed drug discovery and streamline manufacturing. The trend coincides with President Trump’s demands that drugmakers manufacture more drugs on American soil.
Scientists at the Devens facility are using artificial intelligence to discover molecules that can target cancer and other diseases with greater precision. Artificial intelligence can go through data sets from past experiments and identify possibilities that a human might not have considered. Researchers then test these molecules in a virtual world – a process referred to as “in silico”. Only the most promising ones are tested in a physical laboratory. A company can run multiple “in silico” experiments at once.
“Drug discovery and bio-manufacturing are definitely areas where artificial intelligence can have the biggest impact,” said Kyle Chan, a fellow at the John L. Thornton China Center at the Brookings Institution. “These are the areas where artificial intelligence has the greatest advantages over previous approaches due to the need to process and synthesize large and complex data sets.”
Still, there is no guarantee that technological benefits will immediately equal patient benefits. The history of drug development is littered with failures, and it is not known whether molecules identified by artificial intelligence will hold up in clinical trials.
The Bristol Myers Squibb facility is located on an 89-acre campus where the buildings are decorated with portraits of cancer survivors.
Previously, scientists and engineers were never sure why some batches of cells produced large amounts of protein while others failed completely. But now the AI uses information from past batches to identify what variables need to be changed. For example, if the oxygen levels are lower than the previous batch, the system will suggest adding oxygen. If the pH values are higher than the previous batch, it will recommend a correction. It also provides suggestions on the best time to harvest the cells.
These innovations have increased the volume of drugs produced for clinical trials and commercial use at the facility by about 40 percent, according to a company spokesperson.
“We are now able to intervene in batches during the manufacturing process and not have to wait until we get to the end,” said Karin Shanahan, the company’s executive vice president, chief supply chain and chief operating officer.
These innovations helped stabilize the production of Orencia, a drug that treats autoimmune conditions such as rheumatoid arthritis using cells that are extremely difficult to grow. In 2024, production challenges led to shortages in some parts of the world.
The company is just starting to use AI in its manufacturing process for another drug, Breyanza, which turns a cancer patient’s own white blood cells into a personalized therapy. Currently, the Devens plant is approved by the Food and Drug Administration to manufacture the treatment for only 12 patients at a time.
Ms Shanahan said she hoped AI would eventually increase the production of treatments that are often considered a last resort for people with blood cancers such as leukaemia.
Bristol Myers Squibb has embarked on a series of cost-cutting measures as a key patent on its Opdivo cancer drug expires in 2028. The drug, which uses proteins that have been genetically engineered to target cancer cells, generated more than $10 billion of the company’s $48 billion in sales last year.
The company is cutting costs by $2 billion by the end of 2027, on top of $1.5 billion in cuts announced in 2024. More than 1,000 positions are being cut, many of them at a research facility in Lawrenceville, New Jersey, raising concerns that artificial intelligence is taking away jobs in the sector.
At the Semafor World Economy Summit last month, Bristol Myers Squibb CEO Chris Boerner said the company had a responsibility to use AI to support its mission, but acknowledged that it could adversely affect some employees.
“We’re working together with those employees to make them more marketable in this technology — within the company or elsewhere,” he said.
The Devens facility, which was completed in 2009 at a cost of $750 million, was not designed with artificial intelligence in mind. Even in 2020, employees used Excel spreadsheets for some tasks. Batch records documenting each step of production were completed by hand. In recent years, however, the company has prioritized digitization and automation of its processes.
“We needed to make sure that we could formulate our products faster, that we could scale them commercially faster,” Ms. Shanahan said. “And that’s really what made us go down this path.”
Overall, the company is looking to reduce the time it takes to bring a drug to market to about six years, down from nine, she said.
Other factories receiving recognition from the World Economic Forum this year included Yueda Textile in Yancheng, China, which collects sensor data to detect machine maintenance problems before they occur, reducing costs; and Midea, a microwave and air conditioner maker in Thailand that uses AI to investigate customer complaints and generate recommendations for corrective actions that cut resolution times from months to days.
Some of the factories in China that won awards in previous years belong to American companies, including Johnson & Johnson and Agilent, a California-based supplier of high-end laboratory equipment. However, Chinese drugmakers are also expanding their use of AI.
“This is not just a trend among American pharmaceutical firms,” said Mr. Chan, a fellow at the Brookings Institution. “China’s biotech industry is rapidly moving to use artificial intelligence to accelerate progress.”
