Why Ireland needs its own quantum machine — sovereignty, talent, climate

Every other week someone in Dublin asks me whether Ireland really needs a quantum computer of its own when AWS Braket, IBM Quantum and Azure Quantum will rent us cycles by the minute. The honest answer is that renting compute from another jurisdiction is fine for benchmarking and education, and it is the wrong answer for sovereign science, sensitive industrial chemistry, and the next twenty years of Irish technical talent. We are building Ireland Quantum 100 in Co. Tipperary because the country that owns the cryostat owns the queue, and the country that owns the queue decides which problems get solved first.

What "sovereign quantum" actually means

The phrase gets thrown around loosely, so let me pin it down. Sovereign quantum, in the Irish context, means a physical machine on Irish soil, operated under Irish and EU law, where the queue priority, the calibration data, the pulse-level access and the circuit logs do not leave the jurisdiction unless we choose to export them. It is not about distrusting hyperscalers. It is about the simple observation that when a transmon-based quantum processor sits inside a dilution refrigerator at sub-15 mK in Tipperary, the people deciding whose carbon-capture catalyst gets simulated this Tuesday are people accountable to Irish institutions.

That matters more than it sounds. On a current-generation superconducting machine, queue depth is the binding constraint. A variational quantum eigensolver run on a 100-qubit heavy-hex device for a realistic catalyst Hamiltonian is not one circuit — it is thousands of parameterised circuits across an optimiser loop, often with mid-circuit measurement and dynamic decoupling sequences between layers. If you are forty-third in a global queue behind a hedge fund stress-testing a portfolio model, your chemistry run takes a fortnight instead of an afternoon. Ownership of the queue is ownership of the science.

The Irish quantum case: why now, why here

The Irish quantum case rests on three things that have quietly come together in the last eighteen months. First, transmon fabrication has matured to the point where 100-physical-qubit modules with median two-qubit gate fidelities in the 99.5–99.7% range are no longer research artefacts — they are buildable systems with known supply chains for the Josephson junction wafers, the TWPA amplifiers, the coaxial readout lines and the mixing-chamber plates. Second, the dilution refrigerator market has stabilised: Bluefors, Oxford Instruments and a handful of others ship cryostats with cooling power and wiring counts sufficient for a 100-qubit device with room to grow toward distance-3 surface-code patches. Third, and most importantly for Ireland, the SDK ecosystem — Qiskit, PennyLane, Cirq, and the OpenQASM 3 standard — has converged enough that Irish researchers no longer need to learn a vendor-specific dialect to write useful circuits.

That convergence is what makes a sovereign Irish machine practical now in a way it was not in 2020. The hardware is buildable, the cryogenics are procurable, and the software stack is portable. What was missing was the will to do it on this island.

The talent argument is the one most people miss

Quantum talent ireland is the search term, but the reality is messier than the term implies. Ireland has produced quantum physicists for decades — Tyndall, Trinity, UCD, Maynooth, the IPIC photonics work, the cold-atom groups. What we have not produced is quantum engineers in the operational sense: people who can recalibrate a qubit's π/2 pulse at three in the morning because thermal drift moved the resonance frequency, people who can read a randomised benchmarking decay curve and tell you which crosstalk channel is the culprit, people who can write a Qiskit transpiler pass that respects the heavy-hex coupling map while minimising SWAP overhead.

You cannot train those people on cloud access alone. Cloud access teaches you to write circuits. It does not teach you what happens when a TWPA saturates, what a two-level-system defect looks like in a T1 measurement, or how to debug a control-electronics timing skew that is corrupting your mid-circuit measurement. Those skills are learned in the room with the fridge, by people who are paid to be there for years. A sovereign machine creates that room. Without it, every Irish PhD with operational instincts ends up in Zurich, Delft, Chicago or Sydney, and we re-import them at consultant rates a decade later.

This is why the Ireland Quantum 100 programme is structured as much around resident engineers and rotating postdocs as around customer compute time. The machine is the excuse; the cohort is the deliverable.

Climate workloads are not a marketing choice

We have been explicit that climate-science workloads get first-cohort priority on the machine. That is not a press-release decision — it is a technical one. The problems that current noisy-intermediate-scale quantum hardware can plausibly help with in a useful timeframe are dominated by quantum chemistry: ground-state energy estimation for small-to-medium molecules, reaction-pathway analysis where classical density functional theory either fails or gets too expensive, and materials discovery where the electronic structure has strong correlation that classical methods handle badly.

That list maps almost directly onto the climate stack. Carbon-capture chemistry — particularly the amine and metal-organic framework families — is exactly the kind of strongly correlated electronic-structure problem where a 100-qubit transmon device running VQE or quantum-selected configuration interaction can produce results that are at least comparable to, and sometimes better than, the classical CCSD(T) baseline. Photovoltaic absorber discovery, battery cathode chemistry, and nitrogen-fixation catalysts sit in the same regime. Grid optimisation and climate-finance portfolio problems map onto QAOA and quantum-enhanced sampling, which a 100-qubit machine can run at meaningful problem sizes.

For more on how the chemistry pipeline integrates with carbon-offset supplier evaluation, see the climate workloads cohort overview. The short version: the same simulation that tells you whether a candidate MOF binds CO₂ at 400 ppm under flue-gas conditions tells the offset-stack side of the business whether to back the supplier deploying it.

What 100 physical qubits actually buys you

I want to be careful here, because there is a lot of nonsense talked about qubit counts. One hundred physical transmon qubits on a heavy-hex topology, with two-qubit fidelities in the high 99s and reasonable readout, is not a fault-tolerant machine. You do not get a logical qubit out of it under a distance-5 surface code with margin. What you do get is:

• Variational algorithms (VQE, QAOA) at problem sizes that exceed brute-force classical simulation, which means real benchmarking against classical heuristics rather than against exact answers.
• Quantum-selected configuration interaction and sample-based quantum diagonalisation methods that can attack chemistry problems in the 30–60 spin-orbital range usefully.
• Error-mitigation techniques — zero-noise extrapolation, probabilistic error cancellation, dynamical decoupling — at scales where they actually shift the bias-variance trade-off rather than just adding overhead.
• A real platform for the surface-code roadmap: distance-3 patches are within reach on a subset of the device, which is how you build the operational muscle for the distance-5 and distance-7 work that comes with the next generation.

That is the honest envelope. It is not "solve climate change by Tuesday". It is "do chemistry that the classical stack cannot do cleanly, train a generation of engineers on a real machine, and have the surface-code experience in-house when the qubit counts step up". You can read more about the broader programme on the Ireland Quantum 100 page.

Sovereignty is also about what you don't have to ask permission for

One last point that does not get enough air. When a pharma company, a state utility, or a defence-adjacent research group wants to run a sensitive workload on quantum hardware, the current options require them to ship the circuit — and therefore the structure of the problem — to a cloud endpoint owned by a US or Chinese entity. For most workloads that is fine. For some, it is not, and those workloads currently do not get run at all, or get run at a level of abstraction that destroys their value.

A sovereign Irish machine, operated under Irish data-protection and EU export-control regimes, with on-prem submission options for cleared workloads, opens a category of problem that simply has no home today. That is the quiet half of the sovereignty argument, and it is the half that actually pays for the cryostat over a ten-year horizon.

Where to start this week

If you are an Irish researcher, engineer, or technical founder reading this and wondering where to plug in: install Qiskit or PennyLane, port one real problem from your domain into a parameterised circuit, and run it against a free-tier simulator at 20–30 qubits. That single exercise will tell you more about whether quantum is relevant to your work than any white paper will, and it puts you in the small group of people who will be ready to use the Tipperary machine the week the queue opens. The cohort is being built now, not in 2027.