Industry · Early Pilots
Quantum Computing for Pharmaceuticals and Drug Discovery
Drug discovery is expensive, slow, and failure-prone, and part of the reason is that the chemistry of how drugs bind and react is hard to compute accurately. Quantum computers are expected to help with the hardest molecular simulations, but the evidence so far comes from small pilots, and strong AI tools are changing what needs to be solved.
Maturity
Early pilots on small molecules and hybrid workflows
Most credible quantum workload
Accurate electronic structure for hard targets such as metalloenzymes and reactive intermediates
Hardware needed
Fault-tolerant machines; estimates reach millions of physical qubits for cytochrome P450
Strongest classical competition
AlphaFold-class structure prediction, DFT, free-energy methods, and machine-learned potentials
Notable collaborations
Boehringer Ingelheim with Google, AstraZeneca with IonQ, AWS, and NVIDIA, Moderna with IBM
Realistic horizon
Early 2030s for first chemistry results; drug discovery impact is indirect and later
Overview
Bringing a new drug to market commonly takes ten to fifteen years and costs over a billion dollars once failures are counted, and roughly nine in ten candidates that enter clinical trials never reach patients. Much of that failure traces back to biology that was not predicted correctly, but some of it comes from chemistry: molecules that bind less strongly than expected, are metabolized too fast, or carry unforeseen toxicity. Pharmaceutical companies therefore invest heavily in computational chemistry, and many of them treat quantum computing as a possible future extension of that toolkit.
The relationship between quantum computing and drug discovery is often overstated. A quantum computer will not read a disease and output a cure. At most it offers a more accurate way to calculate certain molecular properties, and only for the cases where today's approximations break down. Because AI structure prediction and machine-learned force fields have advanced so quickly, the room for quantum advantage is narrower than it appeared a decade ago, which makes honest scoping especially important.
Why this industry is a candidate for quantum computing
Molecules are quantum systems. The strength with which a drug binds to its target, the pathway by which an enzyme breaks a drug down, and the mechanism of a reaction in a synthesis all depend on the behavior of electrons, which is governed by quantum mechanics. Standard methods such as density functional theory and molecular mechanics approximate that behavior and work well for many problems, and for others, especially systems with transition metals or many nearly degenerate electronic states, the approximations are unreliable.
A fault-tolerant quantum computer could in principle compute the electronic energy of such systems to a controlled accuracy, which removes a source of uncertainty in the pipeline. The remaining challenge is scale: the molecules that matter for drug design are large, and the hardware needed for them is far beyond today's devices. That is why the sector watches the field closely while keeping its day-to-day work on classical and AI methods.
Key facts and numbers
| Item | Detail |
|---|---|
| Development time and cost | Commonly cited as 10 to 15 years and over a billion dollars per approved drug |
| Attrition | About 90 percent of candidates that enter clinical trials do not reach approval |
| AlphaFold recognition | Demis Hassabis, John Jumper, and David Baker received the 2024 Nobel Prize in Chemistry for protein structure prediction and design |
| Chemical accuracy target | About 1.6 millihartree, roughly 1 kcal per mole, needed to rank binding energies reliably |
| Cytochrome P450 estimate | Published resource estimates point to a few million physical qubits and days of runtime for an accurate active-space simulation |
| AstraZeneca workflow (2025) | IonQ, AWS, and NVIDIA reported about a 20-fold improvement in time-to-solution for a Suzuki-Miyaura reaction simulation workflow, compared with earlier implementations |
Where quantum computing could help
Binding affinity and protein-ligand interactions
The problem: Ranking how tightly candidate molecules bind to a target protein is central to lead optimization. Standard force fields treat atoms as balls and springs and miss electronic effects, while full quantum treatments of the binding site are expensive and sensitive to the method chosen.
The quantum approach: Quantum computers could compute the electronic energy of the active site, treated with a quantum method embedded in a classical model of the rest of the protein, a hybrid called QM/MM. Quantum phase estimation would supply energies with controlled error on a fault-tolerant machine.
Where things stand: Demonstrations treat fragments with a handful of atoms on today's devices, and results match what classical methods already provide. Industrial pilots, such as those announced by pharmaceutical companies with IBM, Google, and Quantinuum, focus on defining the benchmark problems where classical accuracy is insufficient.
Related algorithms: Quantum Phase Estimation (QPE), Variational Quantum Eigensolver (VQE), Quantum Imaginary Time Evolution (QITE)
Drug metabolism and cytochrome P450
The problem: Most drugs are processed by cytochrome P450 enzymes, which contain an iron center whose electronic structure is hard to model. Predicting which metabolites form, and how fast, affects dosing and toxicity.
The quantum approach: Fault-tolerant simulation of the P450 active site with quantum phase estimation, using compressed Hamiltonian representations to reduce cost.
Where things stand: A study by researchers at Google, Boehringer Ingelheim, and Columbia University estimated the resources for a useful P450 simulation and concluded that it requires millions of physical qubits, which has served as a target for hardware roadmaps. Meanwhile, classical methods keep improving, so the size of the gap is itself debated.
Related algorithms: Quantum Phase Estimation (QPE), Linear Combination of Unitaries (LCU), Quantum Signal Processing (QSP)
Reaction mechanisms and process chemistry
The problem: Making a drug candidate at scale requires choosing reactions, catalysts, and conditions. Catalytic cross-coupling reactions, such as the Suzuki-Miyaura reaction widely used in pharmaceutical synthesis, involve transition metals and reactive intermediates.
The quantum approach: Hybrid quantum-classical workflows compute energies for reaction intermediates, and quantum computers may later handle the correlated metal centers directly.
Where things stand: In 2025 AstraZeneca, together with IonQ, AWS, and NVIDIA, reported a quantum-accelerated workflow for a Suzuki-Miyaura reaction, with about a 20-fold improvement in time-to-solution over earlier implementations. The result concerns workflow speed on a modest model of the reaction, and it does not show that a quantum computer can beat the best classical chemistry for this problem.
Related algorithms: Variational Quantum Eigensolver (VQE), Quantum Error Mitigation Techniques, Trotter-Suzuki Decomposition
Machine learning for molecular properties and design
The problem: Predicting solubility, toxicity, and activity from structure, and generating new candidates, are machine learning tasks where data are limited and costly to produce.
The quantum approach: Quantum kernels, quantum neural networks, and quantum generative models have been proposed to learn from small datasets and to represent molecular distributions.
Where things stand: Studies show results comparable to classical models on small benchmark sets. Classical deep learning, including large pretrained models, advances quickly, and a quantum advantage in molecular machine learning has not been shown.
Related algorithms: Quantum Kernel Estimation, Quantum GAN (QGAN), Quantum Support Vector Machine
Who is working on it
- Boehringer Ingelheim, which has collaborated with Google Quantum AI on the cytochrome P450 resource analysis and runs a quantum lab in its research organization.
- AstraZeneca, which has worked with IonQ, AWS, and NVIDIA on quantum-accelerated chemistry workflows.
- Moderna and IBM, which have explored quantum methods for predicting the structure of mRNA, and Merck, Roche, Amgen, and others, which have partnerships with hardware and software firms.
- The Novo Nordisk Foundation and Novo Holdings, which fund the Novo Nordisk Foundation Quantum Computing Programme and invested in Phasecraft's Series B in 2025.
- Quantum software and hardware companies such as Quantinuum, Phasecraft, Pasqal, and Qubit Pharmaceuticals that build chemistry tools aimed at the sector.
Realistic timeline
| Horizon | What to expect |
|---|---|
| Now to 2028 | Hybrid pilots on small molecules, benchmarking against classical chemistry, and workflow integration with HPC and AI. Results are mostly proofs of concept. |
| 2028 to 2032 | Early fault-tolerant machines might compute small active spaces beyond exact classical methods, and the first industrially relevant benchmarks appear. |
| 2032 and later | If large fault-tolerant machines arrive, accurate simulation of metalloenzymes and difficult reaction steps could inform candidate selection. Impact on approval rates is indirect and hard to predict. |
How to judge quantum claims in this sector
Pharma announcements mix three very different claims: that a workflow ran on quantum hardware, that it was faster or cheaper than a previous workflow, and that it did something classical computers cannot do. The first is a technical milestone, the second is a benchmark against a specific baseline, and only the third would be quantum advantage. Most results today fall in the first two categories, which is legitimate progress and should be described that way.
When assessing a claim, ask which molecule, how many qubits, what accuracy, and what the best classical method achieves on the same problem with the same effort. Ask also how much of the calculation was classical, since hybrid workflows often use the quantum processor for a small step. Independent replication and publication in peer-reviewed venues are the signs of a result that deserves weight.
What stands in the way
- Hardware scale: the problems with clear quantum value need thousands of logical qubits.
- Moving baseline: AI and classical methods keep improving, reducing the set of problems that need quantum accuracy.
- Validation: computed energies must translate into better decisions, which are tested through years of experiments and trials.
- Talent: few chemists are trained in quantum algorithms, and few quantum scientists understand drug discovery.
What organizations in this sector should do now
- Identify the specific molecular problems in your pipeline where classical methods give unreliable answers, and quantify the accuracy needed.
- Build a small team with computational chemistry and quantum skills, or partner with academic groups and vendors.
- Run benchmark projects with published, reproducible protocols, and compare against the best classical methods.
- Keep investment proportional to the timeline, and continue to prioritize AI and classical tools that deliver today.
Myths versus reality
- Myth: quantum computers will discover drugs. Reality: they may improve certain calculations, and discovery depends on biology, data, and experiments.
- Myth: AI makes quantum computing irrelevant for pharma. Reality: AI is strong where data exist, and quantum could help where first-principles accuracy is needed, though the gap is narrowing.
- Myth: a faster workflow demonstration proves advantage. Reality: it shows an improvement over a specific implementation, and needs comparison with the best classical approach.
The bottom line
For pharmaceutical companies, quantum computing is a long-term capability to build, not a near-term source of drug candidates. The best preparation is to define concrete benchmark problems, develop in-house expertise, and follow independent verification, while capturing value now from AI and classical simulation.
Frequently asked questions
How could quantum computers help drug discovery?
By calculating the electronic energy of molecules and reaction steps more accurately than classical approximations, especially for systems with metal centers or strong electron correlation. This needs fault-tolerant machines.
Has a quantum computer discovered a drug?
No. Experiments so far simulate small molecules or accelerate parts of workflows, and none has produced a drug candidate.
What did AstraZeneca, IonQ, AWS, and NVIDIA do?
In 2025 they reported a quantum-accelerated computational chemistry workflow for a Suzuki-Miyaura reaction with about a 20-fold improvement in time-to-solution over earlier implementations. It is a workflow benchmark, not proof of quantum advantage.
How many qubits would P450 need?
Published estimates indicate a few million physical qubits and days of runtime for an accurate simulation of the cytochrome P450 active site.
Will AlphaFold make quantum computing unnecessary?
AlphaFold predicts protein structure and does not compute reaction energies. Quantum methods target different problems, but AI has narrowed the range of problems where they are needed.
Keep exploring
- Algorithm: Quantum Phase Estimation (QPE)
- Algorithm: Variational Quantum Eigensolver (VQE)
- Algorithm: Quantum Imaginary Time Evolution (QITE)
- Algorithm: Quantum Kernel Estimation
- Hardware: Quantinuum H2
- Hardware: IBM Heron (r1, r2, and r3)
- Hardware: IonQ Forte (and Forte Enterprise)
- Company: Quantinuum
- Company: Phasecraft
- Company: IonQ
- Research: Higher-Accuracy Molecular Energy Estimation Using Adaptive VQE Ansätze
- Research: Automated Active Space Selection Improves Practical VQE Chemistry Simulations
- Industry: Quantum Computing for Healthcare
- Industry: Quantum Computing for Biotechnology and Genomics
- Industry: Quantum Computing for the Chemical Industry
- Industry: Quantum Computing for Materials Science
This page is written for educational purposes and summarizes public announcements, published research, and published resource estimates (last reviewed September 30, 2026). Vendor and customer claims have not all been independently verified, timelines are estimates, and nothing here is investment or business advice.