The brute-force problem AI replaces
AI powered drug discovery exists because the traditional way of finding drugs is slow and expensive. For decades the standard method has been essentially brute force: pick a target protein, generate millions of candidate molecules, and test them one by one hoping something binds. Insiders describe it as throwing spaghetti at a wall and seeing what sticks. Most candidates fail, a single drug can cost more than $2 billion to develop, and the timeline runs a decade or more.
The AI approach changes the first step. Instead of screening randomly, a model learns the patterns that make a molecule bind to a specific target, then searches a vastly larger space of possible designs in a fraction of the time. The speedup is real, and it is concentrated exactly where the old method was slowest. Our AI drug discovery explainer walks through the history in more detail.
Turning proteins into data
The company behind the “sub-$10 AI drug stock” promo, Alexander Green’s Oxford Club pitch for Absci, illustrates the mechanism in its purest form. Absci’s approach is to take proteins, the biological machines that do most of the work in the body, and convert them into a digital representation that a generative model can analyze.
Once a protein is data, the AI can do what no lab bench can: search through billions of possible designs and rank the ones most likely to work before anything is synthesized. Instead of physically testing thousands of compounds, the model ranks designs in silico, and only the most promising handful ever reach a lab. That is a step change in throughput, and it is the specific claim the promo is really making. The framing, a company that will one day simply collect royalties while partners do the rest, compresses the timeline, but the underlying mechanism is sound.
Where the speedup stops
The important caveat is that AI speeds up discovery, not development. A model can propose a promising drug candidate in weeks, but that candidate still has to pass through the same clinical trials as any other drug, and trials take years and fail often. The front of the pipeline got dramatically faster; the back of the pipeline moves at the same regulatory pace it always has.
That is the single most important thing to understand about AI powered drug discovery, and it is the thing most promos leave out. The discovery engine is genuinely faster, but a faster discovery engine does not shorten a Phase 3 trial or change the odds that a candidate fails in humans. The companies winning here are the ones with candidates already in trials, not just platforms with promise.
The data problem nobody mentions
The models are only as good as the data they are trained on, and protein data is scarce relative to the space the models are asked to search. A generative model needs examples of proteins, their structures, and how they bind to targets to learn the patterns, and that data is expensive to produce. This is why the companies with the best partnerships tend to win: a pharma deal is not just revenue, it is access to the data and the biological expertise that make the model smarter over time.
That feedback loop, where each partnership improves the model, is the real moat in this space, and it is more durable than any single drug candidate. It is also why the partnership list matters as much as the pipeline. Merck, AstraZeneca, and Eli Lilly are not just paying customers for Absci; they are sources of the very data that makes the platform valuable.
What it means for investors
For an investor, the practical question is which companies have turned the mechanism into a clinical asset. Absci is a useful example because it has done exactly that: its lead asset is an antibody for pattern hair loss with proof-of-concept data expected in the second half of 2026. The distinction between a platform and a pipeline is the whole ballgame, and it is visible in whether a company has a drug candidate in human testing.
The broader point is that AI powered drug discovery is a real mechanism, not a slogan, and it is worth understanding on its own terms before attaching it to any stock. The clinical asset is the difference between a research project and an investment, and it is the first thing to look for in any name in this space. We cover the generative models that do the actual designing in our generative AI drug discovery explainer.
Ready to see the research? Click here to access Alexander Green’s report.
NewsletterVetter is an independent publication. We receive compensation from some of the services we review through affiliate links. Nothing on this site is investment advice. Always do your own research.