Scientist pipetting sample into tray for DNA testing in laboratory.jpg

Featured image from Getty Images

VC

AI biotechs fetch big premiums as investors pile into drug discovery startups

Founders say they’re workshopping their edge over conventional LLMs for drug development purposes.

As the technological capabilities powering AI-assisted drug discovery advance, startups in the sector are scrambling to explain their edge over conventional LLMs.

Over the last 12 months, VC invested a total of $3.2 billion across 135 AI drug development startups, according to a new PitchBook analyst note. And investors have been paying up to do so. The new research found that AI-native biotech companies have recently been fetching a nearly 100% valuation premium over the biopharma valuations more broadly.

But in their fundraising efforts, founders say one tall task is explaining to investors the technicalities of how their models differ from off-the-shelf, user-driven LLMs like OpenAI’s ChatGPT. An investor may specialize in deep tech, or a VC may have a healthcare background, but it’s rare to be conversant on both fronts.

“There’s just not a lot of people who understand both the tech and the problems it’s trying to solve,” said Brian Alexander, CEO of AI drug discovery and development startup Valo Health.

Biotech startups argue that chatbots inherently don’t meet the needs of drug development, especially when it comes to precisely analyzing 3D medical imaging and evaluating clinical trial results.

Chatbots like ChatGPT are fed gargantuan amounts of data, and they work by synthesizing and predicting sentence structures. LLMs are trained with large amounts of publicly available data, ranging from books to social media. Models become more advanced over time as they interact with human users, but they still rely on predefined data sets to analyze existing information.

Startups are working to create biological models that synthesize all available scientific research and genome datasets. Founders say that’s the relatively straightforward part. The hard part is developing the tech to conduct real-life experiments and run complex clinical trials with AI.

Biotech AI models rely on narrower data sets, like the 20,000 to 25,000 estimated genes that make up DNA. These specialized models have been fine-tuned to run medical simulations with precise imaging capabilities that traditional LLMs lack.

“If you think of ChatGPT as the entire internet, some of these models run on less than a million structures,” said Ben Riccio, an associate analyst at PitchBook.

Latest Research

European VC Valuations Report

The Q3 2026 European VC Valuations Report delves into the sizes and valuations of European venture deals.

“Taking off-the-shelves LLMs and asking them to predict 3D drug protein structures is kind of a futile exercise,” said Evan Feinberg, founder and CEO of Genesis Molecular AI.

Founders say it’s too early to tell whether they’ll upend the traditional and time-intensive clinical trial model. But deals being struck by top GPs and corporate venture capital firms indicate they’re already shaking up a process that has remained relatively stagnant for years.

Recent News

Apollo views daily pricing as key to expanding alternatives access. Industry peers are skeptical.

Daily pricing is expensive, bureaucratic and risks giving individual investors a false sense of what their holdings are worth, critics say.

Slush investors point to hardware and deep tech for defensibility

Investors at Slush say AI is narrowing the gap between startups and incumbents.

Riccio said it’s an opportunity for big pharma companies to make strategic acquisitions.

“It’s been one-off research, heavily siloed stuff—there’s not one, massively and publicly available data set,” Riccio said. “[Large pharma companies] are going to all want to bring this in-house.”

Featured image from Getty Images

Learn about our editorial standards.