Is quantum computing useful for weather and climate forecasting?

No. As of October 23, 2026, Rosetta Quantum counts zero operational weather or climate forecasts running on quantum hardware, under three declared criteria: (1) the forecast is produced operationally — on a schedule, for real users, by a weather service or forecasting center; (2) a quantum processor sits somewhere in the production loop; (3) the system is publicly documented. Every quantum-weather effort we could find is a proposal, an exploratory collaboration, or a simulator study. The obstacles are not engineering details — they are structural, and the most honest review of the field was written by atmospheric physicists themselves. This verdict sits in the ledger next to our map of what quantum computing is used for today (what is quantum computing used for today).

Who runs the world's weather forecasts today?

Classical supercomputers — and, since 2025, operational AI models that raised the bar quantum would have to clear. ECMWF, the European center whose medium-range forecasts are a global reference, runs its physics-based Integrated Forecasting System (IFS) at 9 km global grid spacing, and on February 25, 2025 made its machine-learning model AIFS fully operational — the first operational open ML weather model, with gains of up to 20% on some headline scores including tropical cyclone tracks, at a 28 km grid, using roughly 1,000 times less energy per forecast than the physics pipeline (ECMWF, 2025-02-25).

That last number is the strategic fact most quantum-for-weather pitches skip. The disruption classical AI delivered to forecasting is already here, measured, and in production. Any quantum proposal now competes not against an expensive supercomputer run, but against a GPU model that is three orders of magnitude cheaper in energy and already operational. ECMWF's own June 2026 article on the computing power behind its forecasts discusses HPC and "AI forecasts driven by GPUs" — and does not mention quantum computing once (ECMWF, 2026-06-26).

Why are weather equations hard for a quantum computer?

Because the two things a forecast needs most — reading the full answer out, and nonlinearity — are the two things quantum computers handle worst. The reference analysis is Tennie & Palmer (University of Oxford), “Quantum Computers for Weather and Climate Prediction: The Good, the Bad and the Noisy” (arXiv:2210.17460, 2022-10-31). Three findings carry the verdict:

Readout. A quantum state could hold a 10⁹-variable atmospheric state in about 30 qubits — input is cheap. But measurement returns roughly one bit per qubit, and a forecast center needs the entire field back, every cycle. Tennie & Palmer call the output problem an “(insurmountable) bottleneck” for operational centers (arXiv:2210.17460). This is the same fine print that killed a decade of HHL-based speedup claims — we documented it in what is HHL and why did its fine print kill so many claims.

Nonlinearity. The Schrödinger equation is linear; the atmosphere is not. The workaround — mean-field coupling across N copies of the state — accumulates errors, and chaotic systems with positive Lyapunov exponents (which is what weather is) may need exponentially many copies (arXiv:2210.17460).

Scale. The one concrete gate count in the literature is brutal: Scherer et al. (2017) estimated about 10²⁹ gates to run a quantum linear-system solve of dimension ≈3×10⁹ — forecast scale. At an assumed 10⁻⁹ s per gate that is ≈10²⁰ seconds of runtime, which is about three trillion years (our arithmetic on the published figure), and it would demand gate fidelities around 1−10⁻²⁹ (as analyzed in arXiv:2210.17460). No published shortcut has removed this barrier as of October 23, 2026.

What is actually being researched?

Proposals and simulator studies — dated, and none with a hardware result on real forecast data. The record:

Pasqal and BASF announced in August 2022 an exploratory collaboration to study quantum neural networks for weather-pattern prediction on neutral-atom processors; BASF's own scientist framed the benefit as arriving “once quantum hardware matures,” and no experimental results were reported (Meteorological Technology International, 2022-08-17). The Open Quantum Institute, hosted at CERN for its 2024–2026 pilot phase, lists a weather-and-climate project with ECMWF, Planqc, the University of Oxford and the WMO as contributors — at the stage “Phase 2 – Full Proposal,” proposing quantum and quantum-inspired tensor-network methods “with the potential of achieving a speed-up”; no runs, no results (OQI project page, accessed 2026-10-23).

The most instructive measured data point is honest about being negative: Ueno & Miura (University of Tokyo) implemented a quantum algorithm for a stochastic multicloud parameterization — the subgrid cloud statistics inside climate models — entirely on a Qiskit simulator, and found their quantum method needed about 4 times more samples than classical Monte Carlo to reach the same result (C_Q/C_MC = 4), with circuit depths around O(1000) and, by their own account, no use of superposition that would change the scaling (arXiv:2406.11350). That is what a real research frontier looks like: small, dated, and currently losing to its classical baseline.

Will quantum computing ever help forecasting?

Possibly — but in niches, and everyone serious says so in the future tense. Tennie & Palmer's own conclusion is that quantum computers will be unlikely to replace classical machines for numerical weather prediction in the foreseeable future, and that the plausible role will be specialized subtasks inside a heterogeneous pipeline, the way GPUs complement CPUs today (arXiv:2210.17460). The candidates that will be tested: stochastic subgrid parameterizations of the multicloud kind (where the UTokyo result sets today's honest baseline at 4× worse than classical), and the OQI route of tensor-network methods — which may well end up delivering their speedup as purely classical, quantum-inspired algorithms, a pattern this ledger has documented before in can quantum computing accelerate generative AI. Rosetta will re-verify this verdict when any of those projects publishes a measured run.

What we don't know

We don't know whether corporate or government groups hold unpublished quantum-weather results; our zero counts only what is publicly documented. We don't know whether the OQI proposal will produce a measured pilot before the CERN-hosted phase ends in 2026, or what its tensor-network benchmark will show — including whether the win, if any, will be quantum at all. We don't know whether Tennie & Palmer's speculative idea that hardware noise could be repurposed as the stochastic term in parameterizations will survive contact with real devices — it has not been demonstrated. And we don't know whether future algorithmic work (e.g., on linearization techniques) will cut the copy overhead for nonlinear dynamics; the 10²⁹-gate figure is a 2017 estimate for one algorithmic route, not a law of nature.

Sources

• Tennie & Palmer — “Quantum Computers for Weather and Climate Prediction: The Good, the Bad and the Noisy” (arXiv:2210.17460, 2022-10-31)
• ECMWF — “ECMWF's AI forecasts become operational” (2025-02-25)
• ECMWF — “Computing power is the backbone for sustaining world-leading weather forecasts” (2026-06-26)
• Ueno & Miura — “Quantum Algorithm for a Stochastic Multicloud Model” (arXiv:2406.11350)
• Meteorological Technology International — “Pasqal and BASF to use quantum computing for weather prediction” (2022-08-17)
• Open Quantum Institute (CERN) — Weather and Climate Forecasting project (accessed 2026-10-23)
• Rosetta Quantum — What is HHL and why did its fine print kill so many claims?
• Rosetta Quantum — What is quantum computing used for today?
• Rosetta Quantum — Can quantum computing accelerate generative AI (LLMs)?

Rosetta Quantum publishes verdicts with raw, reproducible data. This is educational content, not a product claim.