A number that changes everything

Josh Baylin, presenting Stansberry Research’s SIR pitch, makes a claim that sounds impossible on its face: AI is “10,000 times faster than even the smartest, most well-funded human scientists on earth” at discovering new medicines.

As we detail in our full review of the presentation, the “10,000 times” figure is a slogan rather than a measured benchmark. But the underlying direction is real, and the speed gap between AI-powered drug discovery and traditional methods is genuinely extraordinary.

To understand why, you need to understand how drug discovery has worked for the last 60 years, and why AI changes the math.

The old way: spaghetti on the wall

The Stansberry presentation includes a remarkable set of quotes from industry insiders. Jonathan O’Connell, who spent 25 years as a drug researcher at GSK and Bristol-Myers Squibb, describes one of the most common methods of drug discovery as “basically a random search. And the idea is to throw as many molecules as you can at it and see what sticks essentially.”

Jeff Liu, who holds a PhD in oncology from Oxford, says “Drug discovery hasn’t really changed in sixty years. Pick a protein, throw millions of molecules at it, spaghetti at the wall, hoping something sticks.”

And biotech CEO Markus Warmuth describes traditional drug discovery as “making millions of molecules and see what sticks.”

This is not a critique from outside the industry. These are the people who do the work, describing how the work is done. The method is essentially brute force. Generate millions of candidate molecules, test them against a target protein, and hope that one of them binds in a way that produces a useful effect.

The success rate is staggering in the wrong direction. Josh cites data showing that 99% of drug ideas never lead to a working product. The cost per drug is more than $2 billion, and the timeline is 10 to 15 years per candidate.

Why AI is different

AI drug discovery does not replace the scientific method. It replaces the brute-force screening step with something smarter. Instead of randomly testing molecules, AI models learn the patterns that make a molecule bind to a specific target. They can screen hundreds of thousands of candidates in hours, not years, and they can prioritize the most promising ones for physical testing.

The MIT example from the Stansberry presentation is illustrative. Researchers wanted to find new antibiotics to address the growing crisis of antibiotic-resistant bacteria. Instead of spending years in a lab, they used AI to screen molecular candidates. Within hours, they had discovered an entirely new class of antibiotic that could destroy some of the most drug-resistant bacteria on the planet.

The presentation notes that the same approach is being used worldwide. Scientists in Seattle used AI to create a new cobra antivenom. A lab in Scotland used AI to sift through more than 4,000 molecules for aging-reversal candidates, identifying 21 promising ones in five minutes, then narrowing to the three best.

The antibiotic resistance crisis

One of the most compelling parts of the Stansberry presentation is its focus on antibiotic resistance, a genuine public health crisis that gets remarkably little attention from investors.

The World Health Organization warns that drug-resistant bacteria could kill 10 million people per year by 2050 if nothing changes. The problem is that bacteria evolve faster than traditional drug development can keep up. If it takes 10 years and $2 billion to develop one new antibiotic, the bacteria will have evolved resistance long before the drug reaches the market.

AI changes this timeline. If AI can screen new antibiotic candidates in hours instead of years, the development cycle shrinks from decades to months. The supply of new drugs can potentially keep pace with the evolution of the bacteria.

This is not a hypothetical. The MIT discovery that Josh describes is a real result, published in real journals, tested against real drug-resistant bacteria. The AI found something that human scientists had not found in decades of searching.

173 drugs and counting

The presentation cites a figure that is worth pausing on: there are more than 173 AI-discovered medicines currently in human trials. These are not lab experiments. They are treatments being tested on real patients.

One is rentosertib, which treats idiopathic pulmonary fibrosis, a disease that causes scarring of lung tissue. For 50 years, no scientist could find a way to reverse it. The AI-discovered drug is now working in human patients at trial sites around the world.

Another is REC-4881, which is being taken by patients with a genetic condition that went decades without a cure. Both drugs were discovered using medical AI, and both are in active human trials.

As Josh notes, these are “treatments that big drug companies, for some reason, couldn’t discover for decades, despite trillions of dollars in research funding. AI has already figured them out in a fraction of the time, for a fraction of the cost.”

The investment implications

The acceleration of drug discovery has implications across the biotech supply chain. If AI can produce drug candidates at 10,000 times the traditional pace, the bottleneck shifts from discovery to manufacturing, testing, and distribution. The companies that handle those steps, the contract manufacturers, the cold-chain shippers, the ingredient suppliers, become the critical infrastructure for the entire industry.

We explore the specific companies in that supply chain in our guide to medical AI stocks, and the broader investment thesis is straightforward: if the pace of drug discovery accelerates, the companies that make the drugs possible see their volumes, and potentially their stock prices, accelerate with it.

The technology is real. The drugs are real. The investment opportunity is in understanding which companies stand to benefit and getting positioned before the broader market catches on.

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.