Climate modelling — what quantum can and can't change in 2027

Where the actual bottleneck sits

Climate models are not bottlenecked on the same thing everywhere. The atmospheric general-circulation core, the ocean general-circulation core, the sea-ice and land-surface couplings, and the sub-grid parameterisations are each a different beast computationally. A useful exercise before talking about quantum is to be honest about which parts a 100-qubit transmon machine in 2027 can plausibly accelerate, which parts it cannot, and which parts won't see a quantum speed-up this decade regardless of qubit count.

Most of the FLOP budget in a CMIP-class run sits in time-stepping the Navier–Stokes-derived primitive equations on a discretised sphere. That is dense, well-conditioned linear algebra at scale. Classical exascale machines do this efficiently and a 100-qubit NISQ-era device does not displace them. That is the unfashionable starting point. Quantum climate modelling in 2027 is not about replacing ECMWF's IFS or NCAR's CESM. It is about attacking the parameterisation closures and the molecular-scale chemistry that those big classical solvers feed off.

What a 100-qubit transmon machine can actually do

Ireland Quantum 100 is a superconducting transmon device on a heavy-hex topology, held at sub-15 mK in a dilution refrigerator, with a surface-code error-correction roadmap that is explicitly post-first-light. At 100 physical qubits with realistic two-qubit gate fidelities in the 99.5–99.8% range and circuit depths bounded by T1/T2 coherence in the hundreds of microseconds, you are firmly in the variational-algorithm regime. That means VQE, QAOA, and the hardware-efficient ansatz family. It does not mean Shor, it does not mean fault-tolerant phase-estimation chemistry, and it does not mean a Harrow–Hassidim–Lloyd shortcut to solving the linear systems inside an atmospheric solver.

What it does mean, concretely, for climate work:

• Small-molecule electronic-structure problems relevant to atmospheric chemistry — radical reactions, ozone-relevant intermediates, OH-driven oxidation pathways — where the active space is 8 to 20 spin-orbitals and a Jordan–Wigner or parity mapping fits comfortably inside the qubit budget.
• Carbon-capture sorbent screening, particularly amine and metal-organic-framework binding-site models where the chemically interesting region can be carved out as a fragment and embedded in a classical mean-field environment.
• QAOA-style combinatorial work on grid-balancing and dispatch problems where the climate relevance is in the energy-system coupling rather than the geophysics directly.

The honest framing: a 100-qubit machine does fragments of the chemistry, well, and lets the classical infrastructure do everything else.

Atmospheric chemistry: the realistic target

Atmospheric quantum work in the near term means treating reaction-rate constants for species that classical DFT handles poorly. The textbook example is multi-reference systems — biradicals, transition-metal-containing intermediates, spin-crossover species — where single-reference coupled-cluster runs into trouble and CASSCF/CASPT2 is expensive. A variational quantum eigensolver with a unitary coupled-cluster ansatz, or more practically a hardware-efficient ansatz with symmetry-preserving gates, can target these active spaces directly.

The output you want for a climate model is not a quantum wavefunction. It is a rate constant or a binding energy, computed once, then handed to a classical kinetic model. This decoupling matters: you do not need the quantum machine in the tight loop of an atmospheric solver. You need it producing better numbers for a lookup table that the classical solver consumes. That is a tractable integration pattern in 2027 and it does not depend on quantum networking, fast classical-quantum round-trips, or fault tolerance.

Ocean and ice: where it doesn't help yet

Ocean quantum simulation is mostly aspirational on a 100-qubit machine. The bulk physics of ocean circulation — thermohaline transport, eddy-resolving dynamics, mixed-layer parameterisations — is large-scale fluid dynamics. The quantum algorithms that look attractive for fluid problems (HHL-style linear-system solvers, quantum lattice-Boltzmann variants, quantum signal processing for differential equations) all assume either fault tolerance, or a state-preparation and read-out regime that destroys the asymptotic speed-up in practice for current hardware.

The same applies to sea-ice rheology and to the cryosphere generally. There is interesting research on quantum-inspired tensor-network methods that can run on classical hardware and may eventually port to quantum devices, but a 2027-era 100-qubit machine is not the right tool. We will be candid with research partners about this rather than over-promising.

The one ocean-relevant niche that is reachable is chemistry-of-the-ocean — carbonate equilibria, dissolved organic matter speciation, calcification chemistry — handled with the same fragment-and-embed strategy as atmospheric chemistry.

The software stack we're building against

Climate simulation quantum work needs an honest software stack. We are targeting OpenQASM 3 as the interchange layer, with Qiskit, PennyLane, and Cirq all supported as front ends. For the chemistry workloads specifically, that means PySCF or Psi4 producing the second-quantised Hamiltonian, OpenFermion handling the fermion-to-qubit mapping, and the variational loop closing through the SDK of the user's choice against our control plane.

For climate-coupling specifically we are building two things: a rate-constant export format that hands clean Arrhenius-fit parameters back to standard atmospheric-chemistry boxes (KPP-compatible), and a job-queue model that treats quantum runs as overnight batch jobs feeding tomorrow's classical run, not as synchronous co-processors. That is the right operational model for a 100-qubit machine in its first year of customer access (Q2 2027 onward).

What we are doing about this

Ireland Quantum 100 is being delivered into Co. Tipperary on a published twelve-month track: site fit-out through Q3 2026, cryostat install in Q4 2026, first-light single-qubit characterisation in Q1 2027, and multi-qubit operation with the climate-priority customer cohort onboarding from Q2 2027. We are deliberately scoping early workloads to atmospheric and carbon-capture chemistry where a 100-qubit transmon device produces results