AI Drug Discovery: A Revolution for the Underdogs

AI won’t revolutionize drug discovery for Big Pharma. They don’t need it.

Pharma has been making remarkable biologic molecules for decades. Not just simple blockers, but a dizzying array of sophisticated therapeutics. Bispecific and multispecific antibodies. T cell engagers. “Masked” molecules that activate specifically in the tumor microenvironment. ATP- and pH-controlled antibodies. Binders to integral membrane proteins like GPCRs and ion channels.

These aren’t lucky accidents—they’re the result of massive long-term investments in research and development that would make most academic departments weep with envy. The global biopharmaceutical industry investment in R&D, across 4,191 public and private companies, was a staggering $276 billion in 2021, according to an analysis in Nature Reviews Drug Discovery.

By comparison, the National Institutes of Health, the biggest source of funding for basic and translational academic research in the US, had a $47 billion budget in 2024.

When a pharma team launches a new drug development program, they build on a basis of shared biological understanding from academic and industry colleagues, and they attack from every angle. The tools in their toolbox are remarkable. Transgenic mice churning out human antibodies. Vast phage and yeast display libraries. High-throughput screening robots testing millions of variants. Computational platforms are useful, sure, but biopharma companies also leverage good old-fashioned medicinal chemistry intuition backed by unlimited resources.

The message is clear: pharma can and will make a drug against any target. For them, AI drug discovery tools are just another arrow in an already overflowing quiver.

The Mirage of Faster, Cheaper Drugs

Breathless headlines promise that AI will slash drug development timelines and costs. As the legendary drug discovery blogger Derek Lowe has repeatedly pointed out, these claims consistently fall flat. Even if we could design perfect binders instantly (which we can’t do—yet), that barely moves the needle on what makes drug development expensive and slow.

Nearly 90% of drug candidates fail in clinical trials despite having preclinical data compelling enough to advance to the next steps in humans. AI doesn’t accelerate manufacturing scale-up. It doesn’t speed up the months of GLP toxicology studies. It doesn’t shortcut the regulatory maze between a promising molecule and a first-in-human trial. The most elegant AI-designed drug still needs to navigate the same treacherous path from bench to bedside.

So if AI won’t make drug development dramatically faster or cheaper for those who already have the tools, what’s the revolution?

The Real Game Changer: Access for Scientists Everywhere

The revolution isn’t about making Big Pharma more efficient. It’s about giving everyone else a seat at the table.

Academic labs and small biotechs are where the truly audacious ideas emerge—the moonshots targeting novel biology that pharma considers too risky. But these groups face a maddening catch-22. Want to work with a premium antibody discovery firm? Prepare to write a check for $200,000 just to start the conversation. Oh, and sign away a slice of your future success through milestone payments and royalty stacks. For most academic researchers living on shoestring budgets, this is a non-starter.

The scarcity mindset, the zero-sum thinking, is literally underpinning the business models for many platform biotechnology companies. They guard their capabilities behind prohibitive paywalls, knowing that academic labs and startups have no other options. The result? The most innovative therapeutic concepts—the ones that could transform how we treat disease—die on the vine because their champions can’t afford the tools to prove them.

AI changes this equation completely.

For over 15 years, I’ve watched computational protein design evolve from the sidelines as a wet-lab protein engineer. The field, pioneered by scientists like Nobel Laureate David Baker, has been laying crucial groundwork, developing the algorithms and principles that would eventually transform drug discovery. But for most of that time, these tools remained the domain of computational specialists. The methods were complex, computationally intensive, and required deep expertise to implement effectively.

That’s changed dramatically. Today’s AI-powered platforms have built on those foundational advances to generate designs with experimental success rates that make high-throughput screening unnecessary. We’ve reached the tipping point where computational design isn’t just theoretically powerful, it’s practically accessible.

I’ve experienced this firsthand in my lab, where these tools have already accelerated our work beyond what I thought possible. The impact was so profound that I founded Ariax Bio to ensure every researcher could access this same power.

At Ariax, we started with BindCraft, a state-of-the-art AI-powered protein design tool developed by Martin Pacesa and Bruno Correia at EPFL. I previously described BindCraft as representing a ‘DeepSeek moment’ in drug discovery: an open-source tool that matches or beats proprietary platforms while remaining completely free to use. Through our web platform, protein design jobs with BindCraft take less than a minute to set up and are completed in hours. No permission needed. No royalties owed. No IP entanglements. Users only pay for compute time—less than AWS charges and orders of magnitude less than they would pay a CRO for conventional protein-discovery approaches.

From Concepts to Compounds

This shift fundamentally rewrites the economics of early-stage drug development. One of the steepest value inflection points in any therapeutic program happens at pharmacologic proof of concept—that magical moment when you prove your hypothesis actually works in a living system.

Previously, reaching this milestone required either massive institutional resources or painful compromises. I’ve watched brilliant researchers shelve transformative ideas simply because they couldn’t access the tools to test them. The gap between “compelling hypothesis” and “testable compound” was too wide to bridge.

Now? An academic lab can go from target identification to designed binders in days. A startup can iterate through multiple approaches for the cost of a single traditional screening campaign. Most crucially, they can walk into investor meetings not with PowerPoint promises but with IND-ready molecules less than a year from the clinic.

This isn’t about competing with pharma on their terms. It’s about changing the terms entirely. As AI leaders like Sam Altman like to say, “You can just do things.”

The Future Belongs to the Bold

Big Pharma has burned through hundreds of billions in R&D dollars by stampeding toward “de-risked” targets. The entire industry chases the same “validated” biology, competing to make incrementally better versions of existing drugs. It’s a rational strategy when you have shareholders to please and billion-dollar infrastructure to feed.

But breakthrough medicines rarely come from playing it safe. They emerge from researchers willing to challenge dogma, to pursue mechanisms others dismiss as too speculative. This is where the underdogs enter the picture. These are the academic labs and small biotech companies unencumbered by institutional inertia. They can move nimbly to seize upon a new technological shift.

AI-powered drug discovery hands these risk-takers the tools they need to prove their wild ideas actually work. We’re entering an era where the limiting factor isn’t the ability to make drugs, but the audacity to imagine new ones. Where a graduate student’s insight can become a clinical candidate without navigating a gauntlet of discovery platform companies and their lawyers.

In my introduction to Ariax, I asked: “What happens when anyone can make a drug?” Let me provide an answer. It means exponentially more attempts at genuine moonshots. A vast expansion of the therapeutic landscape. When barrier-to-entry collapses, conformity collapses with it. Drug discovery is about to get weirder. And that’s exactly what medicine needs.

AI won’t make drugs 10x cheaper or deliver them 10x faster. It will make our drug pipeline 10x more innovative.

At Ariax, we’re building the infrastructure for this future. Not because we think AI will make drug development easy—it won’t. But because we believe the best ideas for new medicines can come from anywhere. And now, finally, anyone can act on them.

 

Aaron Ring is an associate professor in the translational science and therapeutics division at Fred Hutch Cancer Center, and the Anderson Family Endowed Chair in Immunotherapy.