AMD's AI Chief: The Real Breakthrough Is Medicine, Not Chatbots

The man running artificial intelligence at one of the world's biggest chipmakers thinks the technology's defining moment won't be a smarter chatbot. It will be a cure. In a recent CNBC-TV18 conversation on the show Voices From The Valley, Vamsi Boppana, Senior Vice President of the Artificial Intelligence Group at AMD, argued that the most consequential thing AI ever does may happen inside a pharmaceutical lab, not a search box. His striking framing: over the coming decades, AI could help bring more new medicines into existence than humanity has managed in its entire history.

That is a big claim from a man whose company sells the hardware AI runs on. But strip away the optimism and a surprising amount of it is already being proven in clinical trials and peer-reviewed journals. The story of AI drug discovery has quietly moved from promise to patients.

The bet against the chatbot

Most public excitement about AI fixates on text generators, image tools and voice assistants. Boppana's point is that those are the visible surface of something far larger. The same pattern-finding machinery that writes an email can, when pointed at biology, sift through possibilities no human team could ever read in a lifetime.

His logic is essentially about scale. Designing a drug means searching an almost unimaginably vast space of molecules for the rare few that bind the right target, do no harm and survive the body. Humans have explored a tiny sliver of that space across centuries of trial and error. Machines that can model molecules computationally can explore far more of it, far faster. That is the heart of the argument that AI could help create more medicines than people ever have.

Why this isn't just hype

The reason to take the claim seriously is that the foundational breakthrough has already landed — and it won the world's most prestigious science prize. In 2024, the Nobel Prize in Chemistry went to David Baker, Demis Hassabis and John Jumper, recognising computational protein design and AI-driven protein structure prediction.

The tool at the centre of that award, AlphaFold, solved a problem that had stumped biologists for half a century: predicting how a protein folds into its three-dimensional shape from its sequence alone. Shape is everything in biology, because it determines what a protein does and what can switch it off. AlphaFold has since predicted the structures of roughly 200 million proteins — close to every protein scientists have sequenced. That database is freely available and has been used by millions of researchers worldwide.

Think about what that means. For decades, working out a single protein structure could take a researcher years of painstaking lab work. AI compressed much of that to an afternoon and then did it at the scale of nearly all known life. This is the raw material Boppana is talking about when he describes a coming flood of new medicines.

A drug designed by AI is already in patients

The most concrete proof point sits in a clinical trial for a brutal lung disease. Insilico Medicine, a biotech built around generative AI, used its software to do something striking: identify a brand-new biological target, then design a molecule to hit it, largely with machines.

The result, a drug called rentosertib, is a TNIK inhibitor aimed at idiopathic pulmonary fibrosis (IPF), a condition that scars the lungs and is often fatal. In 2025, results from a Phase 2a study were published in Nature Medicine. Patients on the top daily dose saw their lung function, measured as forced vital capacity, improve by an average of 98.4 ml, while the placebo group declined by about 20 ml. For a disease where the goal is usually just to slow the damage, a measured improvement is notable.

It is one mid-stage trial, not a marketed cure, and plenty can still go wrong in later testing. But it is widely described as the first time a drug both discovered and designed with generative AI cleared this kind of clinical hurdle. The proof of concept Boppana is pointing toward isn't theoretical. It has a name.

What actually changes if he's right

The modern drug pipeline is famously slow and brutally expensive. Bringing a single medicine to market typically takes well over a decade and costs sums measured in the billions, with the vast majority of candidates failing along the way. AI promises to attack the two most painful stages: finding a promising molecule and weeding out the ones likely to fail before they soak up years of lab time and money.

If that holds at scale, a few things shift:

• Rare diseases get a second look. Thousands of conditions are too small a market to justify traditional R&D budgets. Cheaper discovery changes that math.
• Timelines compress. Early discovery stages that once ran for years can shrink dramatically, even if human clinical trials still take their own careful time.
• More shots on goal. More candidate drugs entering testing means more chances for the rare success that becomes a real therapy.
• Personalisation becomes plausible. Modelling biology in software opens the door to treatments tuned to specific genetic profiles.

This is also where the longevity thread in Boppana's remarks fits. He isn't promising immortality. The realistic version is more ordinary and more powerful: better drugs for the diseases that shorten lives, arriving faster and reaching more people.

The part a chipmaker can't stay quiet about

There is, of course, a reason an AMD executive is the one making this case. None of it runs on good intentions. Mapping proteins, simulating how molecules behave and training the models behind generative biology all demand staggering amounts of computing power — exactly the AI chips AMD and its rivals build and sell.

Boppana's broader theme in these conversations is that the cost and energy efficiency of computing has to keep improving for this future to be affordable. Every gain in performance per watt lowers the price of running a virtual experiment, which means researchers can run more of them. In that sense the medical revolution he describes is downstream of an unglamorous hardware race. The breakthroughs happen in biology, but they are paid for in silicon.

It's worth keeping the self-interest in view without dismissing the substance. The hardware pitch and the science are both real, and they reinforce each other.

The sober footnotes

A few cautions belong in any honest version of this story. AI can propose a molecule in days, but a human clinical trial still has to prove it is safe and effective, and that remains slow by design and for good reason. Models can also be confidently wrong, generating plausible candidates that fail in living systems. And a richer pipeline raises hard questions about who can afford the medicines that emerge, and how regulators keep pace.

None of that contradicts Boppana's central point. It just sets the timeline. The most important AI breakthrough may indeed be in health rather than chat — but it will arrive as a steady stream of approved drugs over years, not as a single headline. The early evidence, from a Nobel-winning protein atlas to a lung drug already showing results in patients, suggests the direction is set. The open question is only how fast, and how far, it goes.