Quantum Chips Run Colder Than Deep Space. The Gas That Gets Them There Comes From Dismantled Nuclear Weapons β and a NATO Study Just Called It the Industry's Biggest Bottleneck.
To keep a superconducting qubit coherent for even a few hundred microseconds, it has to sit inside a dilution refrigerator chilled to a few thousandths of a degree above absolute zero β colder than any patch of outer space, where the afterglow of the Big Bang still keeps things hovering around 2.7 K. The refrigerant that gets it there, helium-3, isn't mined or drilled. It's a byproduct of tritium β a nuclear-weapons material β slowly decaying, and almost the entire Western supply runs through three specialist manufacturers. In May 2025, a NATO-commissioned study scored the refrigerator itself, not the qubit, as the single most critical bottleneck for two of quantum computing's leading chip designs.
The five-stage trip to 15 millikelvin
Step through the cooling stages inside a real dilution refrigerator, from room temperature down to where the chip actually sits. Node spacing is illustrative, not linear β the real range spans four orders of magnitude, from 300 K to about 0.01 K. Press Next/Prev, drag the slider, or press Run.
The plain version
Inside every serious quantum computer sits a machine that looks a bit like an upside-down gold-plated wedding cake, called a dilution refrigerator. Its job is to get a chip down to about fifteen thousandths of a degree above absolute zero β colder, for what it's worth, than the leftover glow of the Big Bang still drifting through deep space, which sits around 2.7 degrees above that same zero point. At that temperature, superconducting qubits can hold a quantum state long enough to do something useful with it; at room temperature, they can't hold it at all.
Getting there takes a chain of cooling stages, each one colder than the last, finishing with a trick called "dilution": forcing two forms of helium that don't want to mix together anyway, which soaks up heat the same way sweat evaporating cools your skin. The helium doing the work, called helium-3, isn't dug out of the ground. It's a byproduct of tritium β itself a material from nuclear-weapons maintenance β slowly decaying. The US gets essentially its entire domestic supply from servicing its own warhead stockpile, and even that trickle shrinks a little every year the tritium isn't replenished.
Only three serious manufacturers build these refrigerators worldwide, and a NATO-funded study released in 2025 concluded that of everything standing between today's lab prototypes and a useful quantum computer, the refrigerator itself β not the chip β is the single biggest choke point for two of the field's leading designs. A quantum computing boom, it turns out, runs into the same wall a nuclear stockpile does.
The expert version
A dilution refrigerator cools superconducting- and spin-qubit platforms to their operating point (commonly 10β20 mK) through a staged cascade: a closed-cycle pulse-tube cooler brings the system from ~300 K to roughly 50 K and then ~4 K via regenerative gas expansion (no moving parts at the cold end), after which an incoming helium-3/helium-4 mixture begins to liquefy. Below the lambda point, the mixture phase-separates into a helium-3-rich phase floating atop a helium-4-rich superfluid phase; continuously pumping helium-3 vapor off the "still" (~0.7β1 K) drives helium-3 across that phase boundary in the mixing chamber, and because the phase transition is endothermic β helium-3 atoms crossing into the dilute phase absorb heat even as the system approaches absolute zero β the mixing chamber can be held continuously at 10β20 mK, unlike single-shot adiabatic-demagnetization or evaporative-cooling schemes.
The limiting input is helium-3 itself. It's produced almost exclusively as a decay product of tritium β a fusion-boosting and weapons material with a 12.3-year half-life β recovered when the US Department of Energy services aging warheads in the national stockpile; absent fresh tritium reservoirs, the helium-3 yield from existing inventory declines roughly 5.6% annually. Reported bulk prices vary sharply by source β from roughly $1,900β$2,600 per liter to, in one 2026 market estimate, $6,000β$15,000 β against a price near $100/liter as recently as the 2000s; Science's 2025 reporting notes the helium-3 charge alone can account for up to a quarter of the cost of a refrigerator that can otherwise exceed $600,000.
Manufacturing is concentrated in three firms with meaningful Western output β Bluefors (Finland, with US manufacturing in Syracuse, NY), Oxford Instruments (UK) and Janis Research (US) β with Bluefors and Oxford Instruments together estimated at over 70% of global share by one industry analysis. A May 2025 NATO Transatlantic Quantum Community assessment, scoring vulnerabilities across qubit modalities and seven enabling technologies, rated the dilution refrigerator the single most critical bottleneck for superconducting and some spin-qubit architectures β ahead of qubit fabrication itself.
Why it matters for tech + supply chain: if quantum computing scales the way its backers hope, the limit won't be how clever the chip design is β it'll be how much helium-3 a handful of weapons-maintenance programs can spare, and how fast three companies can build refrigerators to hold it.
Why it matters for tech + supply chain: capacity planning for superconducting-qubit roadmaps should treat dilution-refrigerator and helium-3 lead times as a hard constraint on deployment schedules, not a line item β this procurement chain competes directly with national tritium-reservoir maintenance cycles, a supply nobody in the quantum industry controls.