The Tipperary site — why here, what's installed, what stays out

Why Tipperary, and why Clonmel specifically

The site question for a 100-qubit superconducting machine is not "where can we fit a cryostat" — it is "where can we keep one running quietly for a decade". Clonmel sits in a low-vibration inland corridor in south Tipperary, away from coastal seismic noise, away from the heavy-rail lines that thread the east coast, and outside the EM footprint of Dublin's dense cellular and tram infrastructure. The local granite-on-shale geology gives a stiff foundation profile, which matters when your dilution refrigerator's mixing chamber needs to stay below 15 mK and your readout chain is sensitive to anything that wiggles a cable at the millihertz scale.

The other half of the answer is grid and people. Tipperary sits on a stretch of the Irish transmission network with headroom for a sustained industrial draw — a fully loaded dilution fridge, its compressors, its control electronics, and the conditioning plant around them pull a steady hotel load that needs to stay steady. We are also within a 90-minute drive of Cork, Limerick and Waterford, which puts the machine inside reach of the chemistry, materials and climate-science groups we want as the first cohort. A sovereign Irish quantum facility had to be in Ireland's interior, not in a Dublin server hall.

What's actually being installed

The headline hardware is a 100-physical-qubit superconducting transmon processor on a heavy-hex coupling topology, mounted at the mixing chamber stage of a wet dilution refrigerator with sub-15 mK base temperature and roughly 500 µW of cooling power at 100 mK. Heavy-hex is a deliberate choice: it trades qubit connectivity for lower frequency-collision density and cleaner crosstalk behaviour, and it maps directly onto the surface-code and heavy-hex code lattices we expect to need for logical-qubit work later in the decade.

Around the cryostat, the install includes:

• RF control stack — room-temperature AWGs and digitisers driving microwave pulses at 4–8 GHz through attenuated coax, with circulators and TWPAs (travelling-wave parametric amplifiers) on the readout return path for near-quantum-limited amplification.
• Cryogenic wiring plant — NbTi superconducting coax on the cold stages, copper on the warm stages, with thermalisation at every plate. For 100 qubits with frequency-multiplexed readout you are looking at hundreds of physical lines; the cable plant is genuinely most of the engineering.
• Magnetic shielding — nested mu-metal and superconducting shields around the sample stage. Earth's field has to be attenuated to the microtesla range before the qubits will hold coherence.
• Helium recovery — closed-cycle recovery on the pulse tubes plus a buffered helium-4 reserve. Helium logistics are a real constraint on European quantum sites and we are building for it.
• Classical control plane — FPGA-based pulse sequencers with sub-nanosecond timing, a real-time conditional feedback path for mid-circuit measurement, and an OpenQASM 3 / Qiskit / PennyLane / Cirq compatible software layer above it.

What stays out — deliberately

A site like this fails on the things you let in, not the things you install. The Clonmel facility is being built with hard exclusions:

• No co-located GPU farm. Classical accelerators for hybrid workloads sit off-site and connect over dedicated fibre. GPUs are vibration sources, heat sources and EM sources; they do not belong in the same building as a millikelvin stage.
• No shared HVAC with office space. The lab has its own conditioned envelope with vibration-isolated air handling. Office HVAC ducting is a notorious low-frequency noise path into cryostats.
• No wireless inside the shielded zone. Wi-Fi, 4G/5G repeaters, Bluetooth peripherals and DECT phones are excluded from the inner facility. Control traffic is wired.
• No diesel generators on the cryostat slab. Backup power is provided, but the rotating plant is mechanically isolated from the experimental floor.
• No tourist traffic. This is a working facility, not a demo centre. Visits are scheduled and limited because every door cycle is a thermal and acoustic event.

The thermal and EM budget in plain terms

A transmon's coherence time — the T1 and T2 figures that bound how long a quantum state survives — is set by how cleanly you can isolate the qubit from its environment. At 15 mK the thermal photon population at 5 GHz is effectively zero, which is the point. But every cable into the fridge is a potential thermal short and every uncancelled stray field is a dephasing source. The engineering problem is dragging signals in and out without dragging noise in with them.

Practically, that means attenuators staged at 4 K, still, cold plate and mixing chamber to break the room-temperature noise photon flux; infrared-blocking filters in the readout line to kill stray quasiparticle generation; and careful grounding so the whole structure behaves as one Faraday cage rather than a set of antennas. None of this is novel — it is the standard superconducting playbook — but doing it correctly at 100-qubit scale, in a new building, in Ireland, is the work.

Timeline and what each phase actually means on site

The Ireland Quantum 100 delivery runs across the next twelve months on a public schedule:

• Q3 2026 — site fit-out. Slab, shielded room, helium plant, power conditioning, fibre. The boring half of the project and the half that decides whether the next three quarters work.
• Q4 2026 — cryostat install and cooldown. Fridge in, wiring plant in, first cooldown to base temperature. No qubits yet — this phase proves the envelope.
• Q1 2027 — first light, single-qubit. Chip mounted, single-qubit gates characterised, T1 and T2 measured against design.
• Q2 2027 — multi-qubit and first customer access. Two-qubit gate calibration, randomised benchmarking, and the first climate-cohort workloads on the machine.

Next steps

We are publishing the site engineering as we go because Ireland needs a public reference for what a sovereign