How could quantum computers help chemistry?

Last checked: 2 October 2026.

A molecule's behaviour depends on how its electrons share out among their possible orbitals, and the electrons act together, not one at a time. Track every arrangement exactly and the list doubles with each orbital you add (chapter 7). For most molecules chemists use clever shortcuts that work well. They struggle with molecules where many electrons sit close together in energy, often around metal atoms. Those are exactly the molecules in many catalysts and enzymes.

Farms rely on nitrogen fertiliser. Industry makes it with the Haber-Bosch process, at high heat and pressure, using about 2% of the world's energy. Some bacteria do the same job at room temperature, using an enzyme whose core is a cluster of iron, molybdenum and sulfur called FeMoco. Understand how FeMoco works and you might design better catalysts.

In 2017 a team from Microsoft and ETH Zurich worked out the cost of simulating FeMoco on a future error-corrected machine: about 135 logical qubits running for weeks. It became the field's flagship example.

• 2017: 135 logical qubits; about a million physical qubits in one layout, at error rates far better than today's.
• 2021: about 4 million physical qubits at today's error rates, in under four days (Google and others).
• 2022: cytochrome P450, the liver enzyme that breaks down most medicines, about 4.6 million physical qubits for 73 hours.
• 2025: new methods ran 4 to 195 times faster for similar iron-sulfur molecules (Google and others, Physical Review X).
• January 2026: a Caltech-led team reached chemical accuracy on a FeMoco model with ordinary computers (preprint). If it holds, FeMoco stops being the best showcase, and attention moves to larger or stranger molecules.

Small versions. In 2025 IBM and Japan's RIKEN combined a quantum processor with the Fugaku supercomputer to study iron-sulfur clusters, matching the best ordinary approximations. Google and UC Berkeley used a quantum computer alongside a magnetic-resonance experiment to measure molecular structure, matching ordinary results. These are rehearsals for the error-corrected era, not advantages yet.

A 2025 MIT study expects quantum methods to beat the best exact ordinary methods for small and medium molecules in about 10 years, and the cheaper everyday methods in 15 to 20. Company roadmaps aim for machines with hundreds of logical qubits around 2029 (chapter 11). Materials and catalysts look likely to come before medicines.