The cold problem nobody puts on the slide deck
The first thing you learn when you start costing a quantum compute build is that the qubits are the easy part. The cold is the hard part. A superconducting qubit only behaves like a qubit when the thermal noise around it is quieter than the energy gap of the transition you are trying to measure, and that quiet — call it ten millikelvin, give or take — does not exist anywhere in nature. You have to make it. And you have to keep making it, every second, for as long as the machine is meant to run, in a building that has plumbing, lorries passing outside, and a power grid that was not designed with dilution refrigerators in mind. That is the engineering problem. Everything else is software.
I want to walk through what the cryogenics stack actually looks like for the Ireland Quantum facility we have committed to deliver in Q2 2027, because there is a lot of public hand-waving about quantum hardware and very little written about the boring middle layer that decides whether a system works at all.
What ten millikelvin really means
Room temperature is about 293 kelvin. Liquid nitrogen sits at 77 K. Liquid helium boils at 4.2 K. The base plate of a working dilution refrigerator runs at roughly 0.01 K — ten thousandths of a degree above absolute zero. The gap between liquid helium and base plate is, in proportional terms, larger than the gap between a summer afternoon in Clonmel and the surface of the sun. You cannot get there with a single technology. You stage it.
A modern dilution refrigerator does this with five or six nested stages. A pulse tube cooler handles the descent to about 4 K using compressed helium gas and clever valve timing, which is also where most of the noise and most of the vibration comes from. Below that, a continuous flow of mixed helium-3 and helium-4 takes over. The two isotopes phase-separate at low temperature, and pulling helium-3 across the phase boundary absorbs heat in the same way evaporation cools your skin, except it works all the way down to the millikelvin floor. There is no other technique that gets you there continuously. There are one-shot adiabatic demagnetisation systems, but for millikelvin operations on a production cadence, dilution is the only adult option.
Choosing a dilution refrigerator class
You do not buy "a dilution refrigerator" the way you buy a server. You specify a cooling power at a target temperature, and then you live with the trade-offs. The number that matters is microwatts of cooling at 100 millikelvin, because that is roughly the budget you have to absorb every signal line, every passive heat leak, and every photon that finds its way down the fridge from the warmer stages above.
For a sovereign Irish facility, the question is not "what is the biggest fridge on the market." It is "what cooling envelope leaves us room to grow the qubit count without ripping out the platform in 2029." The honest answer, for the build profile we are committing to, is a large-bore system with a base-stage cooling power generous enough to support several hundred coaxial signal lines without choking. That choice cascades into everything else: the size of the room, the floor loading, the helium reserve, the compressor footprint, the height of the ceiling because dilution fridges are loaded from above with the inner can hanging like a chandelier.
None of these decisions are exotic. They are just unforgiving. Pick a fridge that is a class too small and you will spend the next three years working around it.
Vibration: the quiet enemy
Pulse tube coolers vibrate. They have to — they work by cycling gas at one or two hertz — and that mechanical thrum couples into the cold stages through every rigid path it can find. In a superconducting qubit experiment, vibration shows up as flux noise, as microphonic pickup on coax, as charge offsets, as drift in the resonator frequencies you spent six weeks calibrating. You cannot eliminate it. You isolate it.
The standard approach has three layers. First, the building. The fridge sits on an inertia block — a slab of concrete poured separately from the surrounding floor, on its own foundation, mechanically decoupled from the rest of the structure. In Clonmel, that means understanding the local geology, which is limestone with a reasonably high water table, and designing the slab to sit on isolators that handle the residual ground motion from road traffic and the occasional agricultural lorry. Second, the fridge frame itself rides on pneumatic isolators with a low natural frequency, so anything above a couple of hertz is attenuated before it reaches the cryostat. Third, inside the fridge, the pulse tube head is mechanically separated from the cold stages by flexible thermal links, usually braided copper or sintered structures that conduct heat well but transmit force poorly.
You measure the result with accelerometers and you tune until you stop seeing the pulse tube fundamental in your qubit data. There is no shortcut. Every site is different. A fridge that works perfectly in a basement at Delft will misbehave on the third floor of a converted office block in any city you care to name.
Electromagnetic shielding from the building inwards
Qubits are antennas. Superconducting qubit cooling buys you nothing if a stray field at the qubit transition frequency is leaking in from a fluorescent ballast two rooms over. Quantum cryogenics and electromagnetic hygiene are the same problem dressed in different overalls.
The shielding stack also goes from the outside in. The building itself wants a shielded enclosure for the lab area — a Faraday cage, properly bonded, with filtered penetrations for power, signal, and helium lines. Inside the room, the fridge has its own mu-metal and superconducting shields nested around the cold stages, because what you care about at the qubit is not just radio-frequency noise but slow magnetic drift from the earth's field as someone parks a van outside. Every cable that enters the fridge gets filtered, attenuated, and thermalised at each stage, which is also where most of the heat load comes from. You are constantly trading signal fidelity against thermal budget. Adding an attenuator at the mixing chamber improves the thermal photon population at the qubit but consumes some of your hard-won cooling power.
This is the part of the stack that is least visible in marketing materials and most decisive in whether the system actually computes anything useful.
What it costs to keep the cold on, in a small Irish town
I will not put a number on the build, because the number depends on the final hardware partner and on procurement decisions that are still live. I can tell you what the operating envelope looks like in plain terms.
- Helium. Helium-3 is rare, expensive, and not produced in Ireland. You buy a charge, you recycle it, and you build the recovery system as a first-class part of the facility, not an afterthought. Helium-4 is easier but still wants on-site liquefaction or a reliable supply contract.
- Power. A single large dilution fridge with its compressors, control electronics, and supporting infrastructure draws steadily, every hour, every day. The compressors run continuously. The lab HVAC runs continuously. There is no idle mode for quantum facility cooling — you either hold base temperature or you spend a day and a half getting back to it.
- People. A cold system that runs without a cryogenics engineer on call is a cold system that will eventually warm up at the worst possible moment. You staff for that.
- Spares. Pulse tube heads, compressor valves, turbo pumps, gauge controllers. You hold spares on site because the freight time from a continental supplier to Tipperary is not the freight time to Munich.
Doing this in Clonmel rather than Dublin or Cork is a deliberate choice. The grid connection is workable, the floor loading on a purpose-built slab is whatever we pour it to be, the ambient acoustic and electromagnetic environment is quieter than any city centre, and the people we want to hire would rather live near the Suir than commute into a capital. The trade is logistics. We accept slightly longer supply lines in exchange for a site that is, by any reasonable measure, a better physical environment for millikelvin work.
What to do this week
If you are an Irish researcher, engineer, or operator who has worked on cryogenic systems, RF, EMC, vibration analysis, or facility-scale helium handling, write to me. The Q2 2027 commitment is firm and the people problem is the one I am spending most of my time on. The hardware is orderable. The shielding is buildable. The slab will be poured. What I want to hear about this week is who in this country has actually held a base temperature for twelve months without losing it, and what they would do differently if they had a clean site, a serious budget, and a founder who has read the manual. That is the conversation I am trying to start.