AI is quickly becoming a central force in drug development, from powering pharma engines to permanently rewiring the capital markets.
An ascendant group of AI-centric companies is ushering in a new era of “fail-fast” drug development, pushing the entire biopharma sector to adopt advanced machine learning as new investors shovel money into these high-tech new drug hunters.
“The real story is the pressure that tech capital puts on biopharma leadership to fix their failure rates,” Tyrone Lam, chief business officer at GATC Health, told BioSpace in an email.
Lam noted that there has been a big influx of tech money into biopharma. These new investors, he added, “fundamentally understand the value of a ‘fail-fast’ process”—a strategy that puts a premium on figuring out quickly where a product might fail.
For many drugmakers, this means infusing AI into their drug design and discovery processes, accelerating timelines and producing better molecules. But Lam believes the industry should go beyond that. “The mandate for AI can’t just be about discovering more molecules faster,” he said.
“The true paradigm shift will be moving risk management from the end of the drug development cycle to the very front,” Lam added. That is, using AI and predictive accuracy tools to de-risk the entire drug development value chain, particularly the initial investment.
In practice, this could entail more thorough and streamlined pre-clinical work and optimizing trial design by selecting more appropriate endpoints and patients, he offered.
But even as tech money further incentivizes pharma to maximize the likelihood of an asset’s success, this new breed of investors nevertheless brings its own set of cons to the drug development sector. “The risk,” Lam said, “is that tech money might overhype discovery velocity and discount the development and regulatory constraints inherent in the process.”
Orr Inbar, CEO of QuantHealth, a company that conducts clinical trial simulations, is more direct about this risk: “Tech money often comes with expectations that don’t fit how drug development actually works,” he told BioSpace over email. Investors that are unfamiliar with drug development are unlikely to know upfront that the process is “inherently slow and unpredictable.”
‘Computation at the core’
AI appears to have become the apple of Big Pharma’s eye in recent months, with many of the industry’s most prominent players—including Merck, Eli Lilly and Bristol Myers Squibb—investing heavily into the technology to build capacity internally.
There is outreach from the machine learning sector, too: In April, AI frontrunner Anthropic named Novartis CEO Vas Narasimhan to its board of directors. Late last month, the tech company launched Claude Science, an AI workbench designed specifically for the life sciences.
Perhaps the strongest signal of AI’s increasingly central role in biopharma came in May, when AI drug hunter Isomorphic Labs raised $2.1 billion in series B funds. This mammoth round, the second-largest in biotech history, came despite the company having no disclosed candidate yet.
The investor excitement around Isomorphic is driven largely by its tech backing—the startup is owned by Google’s parent Alphabet—and an AI-centric drug development engine.
It’s not just Isomorphic. In recent months, companies like Generate:Biomedicines, Parabilis Medicines and NewLimit have similarly commanded impressive investments, driven in large part by their respective AI technologies.
GATC Health’s Lam calls this group of AI-forward drugmakers “tech-bio,” a fitting name given that these companies are “challenging the assumption that biopharma must be science-first and data-second.” They are “building with computation at the core,” he told BioSpace in March.
Indeed, the key distinguishing fact of tech-bio companies is their AI-first approach to discovering and designing drug candidates.
Isomorphic, for example, touts a platform based on the Nobel Prize-winning AlphaFold family of models. AlphaFold can accurately predict protein, DNA and RNA structures, as well as the interactions among them and with other molecules. The company supplements this engine with its so-called “dataverse,” a deep and curated trove of life science data that, in turn, enables Isomorphic to run “massive volumes of in-silico experiments” in parallel.
Isomorphic leverages this approach across different disease areas and treatment modalities—from cancer to immunology, small molecules to biologics.
Also taking an AI-forward approach to drug development is Generate:Biomedicines, which closed a $425 million initial public offering in March, at the time the biggest IPO haul since 2024. True to its name, Generate uses generative AI to “deliberately generate medicines,” aiming to address the most difficult-to-treat diseases, according to its website.
QuantHealth’s Inbar acknowledges that there is currently skepticism toward the type of AI-driven drug development that Isomorphic and Generate do—he calls it “understandable, though I don’t think it’s entirely fair.”
These companies “have shown that AI can design proteins and molecules far faster than a chemist working by hand,” he said.“That has genuinely changed part of the pipeline.”
Another company that puts AI at the core of drug design is NewLimit, a California-based biotech looking to reprogram the epigenome to tackle aging. NewLimit has developed a proprietary model called Ambrosia, which draws from “nature’s languages and human languages” to design payloads that can “make old cells look & act young.”
The biotech has assembled a pipeline led by NLMT1001, an mRNA-based asset that targets liver cells and restores youthful function to the organ, and which is set to enter human trials next year. NewLimit closed a $435 million series C in June, one of this year’s largest.
A new era
Regardless of how good an AI model is, failure will be part of the drug development game—and in pharma that doesn’t necessarily mean a “bad bet,” as Inbar puts it. Seasoned pharma investors understand that studies can take years to complete, and they are comfortable with a certain level of clinical uncertainty and regulatory complexity.
AI can help ease this but not totally eliminate it. “Uncertainty doesn’t disappear, but it becomes something you can reason about more systematically,” he explained. This uncertainty, however, may bare the risks that come with the rise of tech-bio—and with the surge in tech dollars.
Uncomfortable with slow and risky bets, tech investors might instead choose to channel their money into more certain drug programs, Inbar says, which in turn could form something of a feedback mechanism that loops back to the drugmakers themselves. “If investors are expecting quick returns, there’s a risk that companies chase easier targets rather than tackling the diseases where innovation is most needed,” he said.
Regardless of the risks and rewards, the AI wave has opened a new era for biopharma.
“This will be a permanent rewiring of the capital markets, not a temporary bubble,” Lam said, in contrast to the massive influx of money that occurred during the pandemic. The COVID-19 investments, he added, “were an opportunistic reaction to a societal panic.”
Inbar agreed, noting that because the pandemic surge in funding was tied to a specific moment in time, the money “receded when that moment passed.” AI is different because “what’s driving interest now is a genuine shift in what technology can do.
“That doesn’t go away when sentiment changes,” Inbar said.
