People keep asking why we're putting a 100-qubit superconducting machine in Clonmel and not in Dublin. The honest answer is that quantum hardware doesn't care about postcodes, and the things it does care about — vibration floors, electromagnetic hygiene, grid quality, helium logistics, low-cost expansion space, planning timelines — all line up better here than in any Dublin business park I've costed. The capital is a fine place to sell software. It's a poor place to land a dilution refrigerator.
What a transmon machine actually demands from its site
A 100-qubit superconducting transmon system is, at the physical layer, a stack of constraints. You're cooling a chip to under 15 millikelvin in a dilution refrigerator, which means a multi-stage pulse-tube cryostat with a mixture of helium-3 and helium-4 circulating through a continuous cycle. The fridge itself is mechanically noisy at the pulse-tube head and ferociously sensitive everywhere else. Around it you need racks of room-temperature control electronics — arbitrary waveform generators driving microwave pulses in the 4–8 GHz range, low-noise amplifiers, FPGA-based readout, and a control plane that has to keep gate-level timing inside nanoseconds.
The site requirements that fall out of this are unromantic but specific:
- Floor: a vibration profile that doesn't pollute the fridge's mechanical isolation. Ideally a slab on competent ground, not a suspended floor on a podium.
- EMI environment: distance from trams, lifts, large switching loads, and dense mobile-cell infrastructure. Every stray field couples into your readout chain.
- Power quality: clean three-phase, low THD, and headroom for compressors that draw heavily on start-up.
- Thermal load handling: the room itself becomes a thermal battle. The compressors dump kilowatts of heat that has to go somewhere predictable.
- Helium logistics: deliveries, recovery, and a reasonable answer to a global helium supply that has been periodically tight for a decade.
- Expansion footprint: a 100-qubit system is one fridge. The roadmap to surface-code-protected logical qubits is many fridges. You need room to grow horizontally.
Once you write that list down, Dublin starts to look like a difficult site, not an obvious one.
The Dublin problem, stated plainly
Dublin's strengths are talent density, airport access, and a deep software ecosystem. None of those are quantum-hardware blockers, because the talent commutes, the airport is two hours from Clonmel by road, and the software stack runs over the wire.
Dublin's weaknesses, for this specific class of machine, are real. Property is priced for offices, not for industrial cryogenic facilities with reinforced floors and dedicated helium handling. Grid connections in the eastern corridor are increasingly constrained, particularly anywhere downstream of the data-centre cluster. The electromagnetic environment in the city core is dense — DART lines, Luas traction power, and the general hum of high-rise mechanical plant. And the planning timeline for anything that smells like infrastructure is long enough that a twelve-month delivery window becomes implausible before you've signed a lease.
None of this is a complaint about Dublin. It's a recognition that hardware-heavy sovereign infrastructure has different siting logic than a SaaS HQ.
Why Clonmel works for a sovereign quantum site
Clonmel is in south Tipperary, roughly equidistant from Cork, Limerick, and the M50. It sits on competent ground, has industrial floor stock at sane prices, and a grid connection in the region that hasn't yet been saturated by hyperscale build-out. The electromagnetic baseline is quieter than any urban site I've measured. There's no tram network coupling 750 V DC into the soil under your feet, and no twenty-storey mechanical plant next door cycling chillers against your readout chain.
For an ireland quantum location decision, the practical case for Tipperary is:
- Industrial property at a price that lets you build a proper facility, not a compromised one.
- Grid headroom for compressors today and additional cryostats over a multi-year roadmap.
- A vibration floor that doesn't fight the fridge's isolation system.
- Road access for cryogenic deliveries without urban congestion penalties.
- Planning timelines that match a twelve-month delivery, not a five-year masterplan.
- Proximity to the south-west and midlands universities for staff who don't want to live in Dublin 2.
The talent question is the one people press hardest on. The honest answer is that quantum-systems engineers are scarce everywhere, and the people who do this work for a living are used to relocating. The job is the draw, not the postcode. We've found that being in Clonmel rather than Dublin is, if anything, a small recruitment advantage for the senior cohort — they get to do interesting work without paying Dublin rent.
Climate workloads and why physical location still matters
The first cohort on the machine is climate-science workloads: carbon-capture chemistry, photovoltaic and battery materials discovery, climate-relevant protein folding, and grid-optimisation problems. These are variational and quantum-chemistry workloads — VQE, QAOA, and adaptive ansatz methods running through Qiskit, PennyLane, and Cirq against an OpenQASM 3 backend.
You might reasonably ask why physical site matters when the workload is submitted over the network. Three reasons. First, sovereign compute means the machine, the calibration data, and the workload traces sit under Irish jurisdiction on Irish soil — that's the entire point of the Ireland Quantum 100 programme. Second, the climate cohort includes researchers who genuinely benefit from on-site access during calibration windows, particularly for pulse-level work where you're shaping microwave envelopes against a specific qubit's measured response. Third, the integration with our offset-supplier evaluation pipeline at IMPT runs better when the chemistry team can sit in the same building as the control stack during a hard run.
Clonmel makes all three of those easier than Dublin would.
The heavy-hex topology and the room it needs
The 100-qubit chip uses a heavy-hex coupling topology — the same family of layout that has become standard in superconducting transmon systems because it suppresses frequency collisions and gives a clean path to surface-code error correction. Heavy-hex trades some connectivity for much better calibration tractability, which matters enormously when you're trying to keep a hundred qubits inside their target frequency bands without two-level-system defects pulling neighbours into resonance.
What heavy-hex implies for the site is straightforward: the control wiring count scales with qubit count and ancilla count, and the fridge's wiring tree becomes the dominant thermal-load constraint at the mixing chamber. You want a building where you can run cable trays cleanly, where the rack rooms are adjacent to the cryostat hall rather than three corridors away, and where the RF cabling runs are short enough to keep loss budgets sensible. That's a layout problem, and layout problems are easier to solve when you're designing a building around a machine instead of squeezing a machine into an existing office floorplate.
The roadmap from physical qubits to surface-code logical qubits requires more fridges, not just better ones. Each logical qubit at useful code distance consumes a substantial physical-qubit budget. Plan for that on day one or you rebuild the building in year three. We've sized the Clonmel site for the second and third fridge already; for context on how the first machine is being commissioned, the delivery timeline walks through the milestone gates.
Sovereignty, jurisdiction, and the boring stuff that matters
"Sovereign quantum site" sounds grand. In practice it means a small number of unsexy commitments: the hardware is owned in Ireland, the operator entity is Irish, the data — including calibration traces, workload metadata, and result sets — is held under Irish and EU jurisdiction, and the supply chain for critical consumables (helium, replacement components, dilution-unit servicing) has at least one path that doesn't depend on a single transatlantic vendor.
None of those commitments are easier in Dublin than in Clonmel. Several are easier here. Industrial leases are simpler, helium logistics are no worse, and the political and planning environment in a regional town tends to be more constructive about a piece of infrastructure that brings serious technical employment with it.
Where to start this week
If you're a climate researcher, a materials chemist, or a grid-optimisation team trying to work out whether the tipperary quantum machine is relevant to your problem, the useful step this week is to write down your Hamiltonian or your QUBO honestly — qubit count, circuit depth, measurement budget, classical pre- and post-processing — and send it in. We'd rather have a real conversation about whether your workload fits the first cohort than send you a brochure. The machine is being built for problems, not the other way round.