iStock,
ClaudioVentrella
Pfizer, Eli Lilly, Novartis, Bristol Myers Squibb and AstraZeneca are all ramping up the use of AI, but drug discovery is not the primary success story—yet.
How did Pfizer
slash billions
in spending across R&D and administrative operations over the past year? AI of course.
“We didn’t just cut cost, what we did is we improved productivity,” Pfizer CEO Albert Bourla said during a Feb. 3 call to present
fourth quarter earnings
. “And the main lever—of course, there [were] simplification efforts that also took place—but the main lever was the successful deployment of AI.”
Executives detailed the many ways that artificial intelligence has allowed the 177-year-old pharma to streamline operations, presenting case studies that highlight a growing trend among the major pharmaceutical companies.
No longer is AI a niche idea that could help in the future. Investors and other industry watchers expect pharmas to be using the technology today. This was made clear with analyst questions across the pharma earnings calendar about how AI is being deployed and when critical functions like clinical trials might start to speed up with the help of machine learning.
Startups
AI-Focused Biotech Unicorns Face Chilly Market Where IPOs Aren’t Guaranteed
Seven biotech unicorns are advancing AI-powered drug discovery and development—but must contend with a difficult investing environment where competition is steep and the usual roads to exit are uncertain.
July 16, 2025
·
14 min read
·
Tristan Manalac
Read More
“The technology is ready now,” Bourla insisted.
But it’s not delivering exactly in the way many have long hoped—finding new medicines. Rather, the technology is thus far providing comparatively less exciting, but still critical, advances to the drug discovery process.
“The real promise of AI that people were so excited about is that it can uncover new biology,” John Wu, managing director and partner in the health care group at consulting firm BCG, told
BioSpace
. “It can design new drugs that humans otherwise couldn’t do or will take forever to do. I think the proof of that has yet to emerge, but we’re getting closer.”
Lilly Is Past the Leading Edge
In January, Eli Lilly announced a
collaboration with NVIDIA
to develop a co-innovation lab using the famed computer chip maker’s BioNeMo platform and Vera Rubin architecture. The companies agreed to put up to $1 billion into the initiative over five years to help solve some of the key problems of drug discovery using AI.
Another January deal saw
Lilly team up
with China’s Chai Discovery to design novel biologic therapeutics. The biotech is backed by OpenAI, with staff originating from the large language model (LLM) company and Meta’s Fundamental AI Research (FAIR) lab, Stripe and Google X.
In September 2025, Lilly opened up TuneLab, handing over access to its AI suite to select biotechs in exchange for data-sharing to further grow the models that underpin the technology. The initiative is part of Lilly Catalyze360, the pharma’s broader external innovation program designed to nurture the early-stage biotech ecosystem.
Lilly has always invested in small, early biotechs, but until TuneLab, AI was not a specific pillar of Catalyze360, said Aliza Apple, VP of Catalyze360 and global head of Lilly TuneLab, in an interview on the sidelines of the J.P. Morgan Healthcare Conference in January. The AI program represents about $1 billion in R&D spending.
Artificial intelligence
AI-Enabled Clinical Improvements Confirm Biotech Hype as Success Rates Rise
As drug candidates discovered via AI move into later-stage clinical trials, the technology seems to be doing as promised: speeding drug development.
January 21, 2026
·
3 min read
·
Annalee Armstrong
Read More
“We’re not really at the frontier model, leading edge trying to figure out where the next breakthrough in architecture is going to be,” Apple said. “We’re much more focused on what is actually useful today, and how do we make sure that beyond Lilly’s walls, our biotech ecosystem is able to actually take advantage of that as well.”
What has been most useful to Lilly, and what the pharma wants to share via TuneLab, has been the ability to predict outcomes. “We live in a predict-first era already,” Apple said. “That’s no longer aspirational; that’s operational.”
Part of the TuneLab suite is a collection of small molecule and antibody predictive models that the larger pharma runs all of its prospective programs through. The models were created based on decades of research at Lilly.
“We run all of this before we actually ever synthesize a molecule and start running experiments,” Apple said.
Pfizer Wants More Time, Fewer Screens
Pfizer is taking advantage of today’s technological capabilities, with plans to add more than 1,200 graphics processing units (GPUs) in data center investments to support new AI advancements. The technology is now being deployed across every inch of the company, from discovery to legal, manufacturing to marketing, Bourla said on the earnings call.
The proof is in how Pfizer has absorbed numerous companies via M&A recently—Metsera, 3SBio and Seagen among them—which has increased the company’s R&D burden immensely, CFO Dave Denton said. At the same time, the company plans to spend $11 billion on R&D activities for 2026.
“We are able to be more productive in the infrastructure across R&D and take on more substrate to be able to focus on creating medicines for the end of the decade and beyond,” Denton said in response to a question about how Pfizer is integrating AI.
Manufacturing
Drug Manufacturers See Long-Term Opportunity in Climbing AI Use
Companies are moving from using AI for distinct operations to applying the technology for control and optimization of the whole production process.
September 30, 2025
·
4 min read
·
Nick Paul Taylor
Read More
The technology has boosted productivity and freed up dollars to snowball into new medicines. Pfizer is pairing AI engineers with scientists across the R&D organization to measure success and productivity, explained Chief Scientific Officer Chris Boshoff.
“Productivity is speed and cost. We bring costs down by embedding AI and, obviously, accelerating speed,” Boshoff told analysts.
In commercial, Pfizer has used AI to train field forces and make the best use of limited time with physicians who may prescribe the company’s medicines. “We invest more time with physicians rather than behind screens,” said Aamir Malik, chief U.S. commercial officer for Pfizer.
The tech has also helped Pfizer tailor marketing strategies and boost return on investment. This has helped lower selling, informational and administrative expenses, as reported during fourth quarter earnings. The international team has been able to easily adapt materials to different markets that may have different regulatory requirements, according to Chief International Commerical Officer Alexandre de Germay. Whereas before, they would have to individually remake promotional materials to match each jurisdiction’s laws and regulations, now AI can make the change in a snap.
Novartis’ Magic, AstraZeneca’s Speed, BMS’ Challenge
Novartis’ executives do not think AI can do everything, but they are nevertheless throwing down a lot of deals in the space, including multiple partnerships with Microsoft, Google’s Isomorphic Labs and Generate:Biomedicines.
“It’s not a magic panacea,” Fiona Marshall, Novartis’ head of biomedical research, told
BioSpace
at J.P. Morgan. “It’s not like suddenly AI is gonna come up with a drug tomorrow, which is what some people claim—I don’t agree with that at all. It can replace some bench-level science, but not all.”
AstraZeneca, meanwhile, is seeing AI accelerate key parts of the drug development process. “What we are seeing in early discovery, where we have applied AI, is more than 50% faster target drug design and validation,” head of U.S. oncology Mohit Manrao said in an interview at J.P. Morgan.
Artificial intelligence
FDA Rolls Out Agentic Models in Bid to ‘Modernize’ Regulatory Workflows
Agentic AI can help FDA staff manage meetings, conduct pre-market reviews and validate reports, among other tasks, though the agency emphasized that using this technology is optional for its employees.
December 2, 2025
·
1 min read
·
Tristan Manalac
Read More
And when it comes to clinical development, AstraZeneca is using AI as synthetic control arms—virtual control groups based on real-world data—so that the company’s trials are neither “overpowered nor underpowered,” according to Manrao.
Looking for New Drugs
Beyond all of the operational efficiencies, companies big and small are still eager to harness AI to find new medicines, albeit carefully.
“You’ve got to understand the disease and find the target for the disease. That’s one whole area which you can apply AI to,” Novartis’ Marshall said. “You use AI to really combine huge datasets, whether that’s human genetics, clinical studies, tissue samples, real experimental data, historical data.”
Marshall claimed that Novartis has access to the largest corporate database in the biopharma world, called
data42
, which the company probes using in-house developed AI algorithms to discover targets and design new molecules.
Bristol Myers Squibb, on the other hand, is applying AI to a disease that has long defied modern drug development efforts. In October 2025, the pharma doubled down on a
partnership with Insitro
to find novel drugs for amyotrophic lateral sclerosis (ALS). At J.P. Morgan in January, BMS Chief Research Officer Robert Plenge told
BioSpace
that AI is being used in that partnership to go back to basics to understand the pathology behind the disease.
Insitro’s tech is helping to define cellular readouts that are different between the healthy state and the disease state. This is called reversion screening. The goal is to find a way to intervene in the process to avoid or treat the disease.
“We’re beginning to find previously unknown biology that maps to some of the genetic nodes that we’ve actually talked about,” Plenge said.
Artificial Intelligence
The Future of Pharma Is Predictive Care Powered by Consumers, AI: PwC
The world of healthcare is evolving to more predictive care and patients are taking greater control—a trend already emerging around GLP-1 weight loss treatments. As PwC warns, pharma will need to be ready.
September 17, 2025
·
2 min read
·
Annalee Armstrong
Read More
Plenge is excited about the possibilities but pragmatic about the long-term future of AI as science across the biopharma world advances.
“I don’t think there’s going to be this magic unlock in the next 12 to 24 months that suddenly what was impossible becomes possible. But if you take the long-term view, AI is going to be an important part of that.”
The Litmus Test
Like Isomorphic and Microsoft, tech companies are eager to work with pharma companies on AI projects. Besides Lilly, NVIDIA has partnerships with
Novo Nordisk
,
Genentech
,
Johnson & Johnson
and
Amgen
, all aimed at drug discovery.
The tech company also has a hand in earlier biotech investing, including participation in cell therapy company ArsenalBio’s
$325 million series C
in September 2024. That biotech—which has been dubbed a unicorn thanks to a billion-dollar valuation—is
working with BMS
on a handful of CAR T programs that were discovered via an AI-enabled model of the T cell.
Meanwhile, pharma-partnered programs are starting to report key milestones with drugs that were found by AI. Takeda reported in December 2025 that AI-designed molecule zasocitinib
eased the severity
of plaque psoriasis in two late-stage clinical trials. The drug was the result of a pact with Nimbus Therapeutics.
Apple noted that AI is moving so fast that by the time an asset enters the clinic, the model it was created from is probably obsolete. That’s why she wants to make Lilly’s predictive technology part of every workflow, and why the company decided to open access to the biotech ecosystem via TuneLab.
Nevertheless, Apple said she hopes the current slate of clinical AI-created assets succeed for the patients that need them. The goal, of course, is to find new ways to tackle long-intractable diseases.
“There’s interesting data points, but we’ve also seen that the models have improved dramatically over the time that those were in clinical development,” Apple said. “I don’t think of those [early programs] as being the litmus test for whether AI is going to be useful.”
Dan Samorodnitsky contributed reporting to this story.
.responsive-container {
display: flex;
flex-wrap: wrap;
border: 1px solid #ededed;
font-family: Helvetica, Arial, Sans-Serif;
padding: 20px 20px 10px 20px;
}
.column-left {
flex: 1;
max-width: 45%;
padding: 20px 20px 10px 20px;
}
.column-right {
flex: 2;
padding: 20px 20px 10px 20px;
}
@media (max-width: 768px) {
.column-left,
.column-right {
max-width: 100%;
flex: 100%;
padding: 10px 0;
}
}
Subscribe to BioPharm Executive!
Market insights, trending business and policy stories for biopharma leaders
hbspt.forms.create({
region: "na1",
portalId: "4413123",
formId: "7c139b3c-ff0b-44e0-bffe-f449801dca59",
onFormSubmitted: function($form, data) {
window.dataLayer = window.dataLayer || [];
window.dataLayer.push({
'event': 'GTMevent',
'event_name': 'newsletter_sign_up'
});
}
});