Why medicine keeps coming up
Every drug works by molecules interacting: a small molecule slots into a protein, an antibody grabs its target, a reaction pathway runs or doesn't. Those interactions are governed by electrons, and electrons obey quantum mechanics. When a chemist asks "how strongly will this candidate bind?", nature answers with a quantum calculation.
Classical computers can only approximate that calculation. The exact equations blow up exponentially (a molecule of modest size has more electron configurations than any supercomputer can track) so the field runs on clever approximations (density functional theory and friends) that are fast but sometimes wrong in exactly the cases that matter: strongly correlated electrons, transition metals, bond-breaking. A quantum computer speaks the molecule's native language, which is why Richard Feynman proposed the whole idea in 1981: to simulate nature, build something that runs on nature's rules.
The canonical target
The example the field returns to is FeMoco. The iron-molybdenum cofactor at the heart of nitrogenase, the enzyme bacteria use to fix nitrogen at room temperature. Industrial fertilizer production does the same job with the energy-hungry Haber-Bosch process. FeMoco's electronic structure is exactly the kind of strongly correlated quantum problem classical methods struggle with, and a 2017 peer-reviewed analysis showed a future error-corrected quantum computer could tackle it. Turning "how does nature fix nitrogen so cheaply?" into a computable question. That paper, more than any press release, is why chemists take the field seriously.
What drug hunters actually hope for
Nobody serious expects a quantum computer to "discover a drug" end to end. Discovery is a decade-long pipeline: target identification, screening, lead optimization, toxicity, trials. The realistic hope is quantum subroutines inside that pipeline. More accurate binding-energy estimates so fewer false candidates go to the wet lab, better modeling of reaction mechanisms, better predictions for tricky chemistries like metalloenzymes.
The near-term shape is hybrid: classical machines do the heavy screening, and a quantum processor is called for the specific sub-calculation where electron correlation defeats the approximations. Even a small accuracy edge matters when a single failed clinical candidate can cost hundreds of millions.
The realistic state today
Today's machines can simulate only small molecules (sizes classical computers handle easily) so current work is about building methods, benchmarks and teams, not discovering medicines. The honest milestone to watch is a quantum calculation of a pharmaceutically relevant molecule that beats the best classical method on accuracy or cost. It has not happened yet, and researchers who say so out loud are the ones to trust.
What HAS happened is real institutional commitment: major pharmaceutical companies and hospital systems have standing quantum programs, documented below: bets on readiness, placed years before the payoff.
A ressalva honesta: Research demonstrations and hybrid pipelines: not yet proof of broad commercial advantage.