What’s Holding Drug Discovery Back
Drug discovery is increasingly expensive, slow, and uncertain.
Costs have doubled approximately every nine years since 1949, and even with modern tools, the discovery phase remains a major bottleneck. The standard approach—screening existing libraries and iteratively optimizing known compounds through the Design-Make-Test-Analyze (DMTA) cycle—can take years per program.
Despite decades of innovation, only half of discovery programs advance to preclinical development. The problem isn’t that chemical space lacks potential. It’s that our ability to search it is fundamentally constrained.
We’ve Barely Scratched the Surface of Chemical Space
Chemical space is unimaginably large—estimated at 10⁶⁰ drug-like molecules—yet our current methods have explored less than 0.000000001% of it.
This isn’t because we lack the tools to model molecules. It’s because every search begins with what we already know: familiar starting points, well-characterized scaffolds, or pre-built libraries. These approaches are inherently limited. They optimize within the known, rather than discovering the new.
The question is no longer “are there more molecules out there?”
It’s “how many life-saving drugs are still waiting to be discovered—and how do we reach them?”
A Generative Shift in Small Molecule Discovery
At Variational AI, we believe the way forward is not to screen more—but to generate better.
We’ve developed a generative AI foundation model that enables a fundamentally different approach to early discovery. Rather than rely on fixed libraries, we generate novel, diverse, and synthesizable small molecules designed to meet specific therapeutic goals from the outset.
By removing the dependency on known chemical starting points, our approach allows for true exploration—navigating previously unreachable corners of chemical space with speed and precision.
This is not a marginal improvement to the DMTA cycle. It is a step change in how we approach small molecule drug discovery.
New Molecules. New Possibilities.
Our generative-first approach enables the discovery of first-in-class and best-in-class compounds across therapeutic areas—without needing proprietary data, screening campaigns, or costly iterations.
We’ve built this platform to transform how early-stage discovery is done, starting with oncology and expanding across indications.
Because the answers we seek aren’t hidden in bigger datasets or larger libraries.
They’re in the molecules we haven’t made yet.