Boehringer Ingelheim: Quantum Chemistry Partnership with Google Quantum AI

The Problem

Drug discovery fails at a staggering rate. Around 90% of drug candidates that enter clinical trials never reach patients, and a significant fraction of those failures trace back to a molecule behaving differently from what computational models predicted. Better molecular simulation is one of the most direct paths to improving that hit rate.

The core computational challenge is electronic structure: given the positions of the atomic nuclei in a molecule, compute the energy of the lowest-energy arrangement of electrons (the ground state). This energy governs how tightly a candidate drug binds its target protein, how it is metabolised in the body, and whether it forms unwanted side products.

Classical methods divide into two families. Density functional theory (DFT) is fast and scales to thousands of atoms but relies on approximations that break down in strongly correlated electronic environments. Coupled cluster methods, particularly CCSD(T), the “gold standard” of quantum chemistry, are far more accurate but scale steeply with the number of electrons, limiting practical use to relatively small systems. The chemistry that matters most in drug metabolism, such as the transition-metal active sites of enzymes, is often precisely where classical methods struggle.

Why Boehringer Partnered with Google

In January 2021 Boehringer Ingelheim announced a partnership with Google Quantum AI to research and implement quantum computing use cases in pharmaceutical research and development, with a particular focus on molecular dynamics simulations. Boehringer described itself as the first pharmaceutical company worldwide to join forces with Google in quantum computing.

The collaboration is structured as a three-year partnership and is co-led by Boehringer Ingelheim’s newly established Quantum Lab. It combines Boehringer’s expertise in computer-aided drug design and in silico modelling with Google’s resources as a leading developer of quantum computers and algorithms. The stated ambition is to accurately simulate and compare much larger molecules than is currently possible, opening new opportunities for pharmaceutical innovation. On Google’s side, the quantum algorithms group is led by Ryan Babbush.

The partnership was announced as exploratory research into an emerging technology. It did not claim specific simulation results, hardware benchmarks, or a production capability.

Why Quantum Computers Suit This Chemistry

The many-body quantum mechanics of strongly correlated electrons is a natural target for quantum computers because the Hilbert space of N electrons grows exponentially with N, and a quantum computer’s state space also grows exponentially with qubit count. The argument, originally due to Feynman, is that simulating quantum systems is exponentially hard for classical computers but natural for quantum ones.

The Variational Quantum Eigensolver (VQE) is a leading near-term algorithm for molecular ground-state estimation. VQE is hybrid: a quantum circuit prepares a trial wave function, a classical computer measures the expectation value of the molecular Hamiltonian and updates the circuit parameters, and the loop iterates until convergence. VQE uses short circuits and tolerates some noise, which makes it a candidate for today’s NISQ processors, though the variational approximation may not be tight and the classical optimisation is hard.

Setting Up a Molecular Hamiltonian

OpenFermion is an open-source library, developed within Google’s quantum ecosystem and used with Cirq, for turning molecular Hamiltonians into forms suitable for quantum computers. A typical workflow starts from a classical chemistry package (such as PySCF) to compute electron integrals, then encodes the fermionic Hamiltonian as a sum of qubit operators. The snippet below is a simplified illustration of that setup. It is a teaching example, not Boehringer Ingelheim’s code or a reproduction of any result from the partnership.

# Illustrative only: setting up a small molecular Hamiltonian for VQE.
# Teaching example, not Boehringer/Google code or results.
from openfermion import MolecularData, get_fermion_operator
from openfermion.transforms import jordan_wigner
from openfermionpyscf import run_pyscf

# A tiny model system, not a real drug target
geometry = [('H', (0.0, 0.0, 0.0)), ('H', (0.0, 0.0, 0.74))]
molecule = MolecularData(geometry, basis='sto-3g', multiplicity=1, charge=0)
molecule = run_pyscf(molecule, run_scf=True, run_fci=True)

print(f"HF energy:  {molecule.hf_energy:.4f} Ha")
print(f"FCI energy: {molecule.fci_energy:.4f} Ha (exact reference)")

# Map the fermionic Hamiltonian to qubit operators
fermion_h = get_fermion_operator(molecule.get_molecular_hamiltonian())
qubit_h = jordan_wigner(fermion_h)
print(f"Jordan-Wigner Hamiltonian terms: {len(qubit_h.terms)}")

Under the Jordan-Wigner encoding each spin-orbital maps to one qubit. Realistic drug-relevant active sites, such as the iron-containing centres of cytochrome P450 enzymes, require many more qubits than current hardware can run reliably, which is why near-term work focuses on small fragments and on building workflows rather than on producing decisive results.

Active Space Approximation

No quantum or classical method can simulate every electron in a large enzyme. Active space approximation selects the subset of orbitals and electrons that contribute most to the correlation energy relevant for the chemical question being asked, and treats the rest at a cheaper level of theory. Choosing an active space well is part of the craft of quantum chemistry, and it is what makes it possible to map a drug-relevant question onto a small enough problem to study on near-term hardware at all.

Why This Case Matters

Quantum chemistry is widely regarded as one of the most credible long-term applications of quantum computing, because the scaling argument is clearest: simulating N strongly correlated electrons is exponentially hard classically but is expected to be tractable on a sufficiently capable quantum computer. The gap between that promise and today’s hardware is real and large.

The Boehringer Ingelheim and Google Quantum AI partnership is best understood for what it actually is: an early, exploratory, multi-year research collaboration that pairs deep pharmaceutical chemistry expertise with a leading quantum hardware group, aimed at developing the algorithms and know-how that could one day make quantum molecular simulation useful in drug R&D. For students, it illustrates the VQE workflow, the role of active space approximation, and the honest distance between current capability and pharmaceutical relevance.

Learn more: Cirq Reference | OpenFermion Reference

Sources

• Cooperation with Google Quantum AI (Boehringer Ingelheim)
• Partnering with Google in quantum computing (Boehringer Ingelheim)
• Boehringer Ingelheim and Google Partner to Investigate Quantum AI for Pharma Research (The Quantum Insider)