Will Cryogenics Become A Bottleneck for Quantum Computing?

Many of today’s quantum computers, notably superconducting systems, which are the most widely-used commercially at the time of writing, operate only inside dilution refrigerators that depend on a rare isotope called helium-3. In September 2025, Bluefors, whose refrigerators cool much of that hardware, agreed to buy up to 10,000 litres a year between 2028 and 2037 from Interlune, a Seattle start-up planning to harvest helium-3 from the Moon. That annual purchase ceiling is equivalent to a quarter of the 40,000 litres of world output the US Geological Survey (USGS) estimated for 2025.

A supplier does not lock in a decade of lunar gas unless it doubts that terrestrial resources can keep pace. Helium-3 production is small, slow to expand, and shared with security, medicine, and research, while growth forecasts for the quantum computing industry point to many more refrigerators being needed. If quantum computer sales track those forecasts, helium-3 could become one of the industry’s most pressing constraints. Whether this pending bottleneck comes to pass depends on how quickly new quantum computing systems are installed, how much helium-3 each requires for their dilution refrigerators, and how much qualified supply quantum hardware developers can secure.

Why the Coldest Quantum Computers Depend on Helium-3

Superconducting and spin qubits are built into solid chips, where warmth in the surrounding material readily disturbs their fragile quantum states, whereas trapped-ion and neutral-atom machines hold individual atoms in a vacuum and cool them with lasers. Superconducting circuits also conduct without resistance only at extreme cold, so these chips must run far colder than anything in nature.

Such temperatures are measured in kelvin, a scale starting at absolute zero (-273.15°C), and a millikelvin is one-thousandth of a kelvin. These processors typically operate at 10 to 20 millikelvin, up to 270 times colder than deep space. A dilution refrigerator gets to this temperature by circulating helium-3, a light isotope, mixed with ordinary helium-4. Helium-4 forms continuously as uranium and thorium decay in rock and collects in natural gas fields, which yielded around 190 billion litres worldwide in 2025. Below a certain temperature the mixture separates into two phases, and drawing helium-3 across that boundary absorbs heat, resulting in the cooling effect.

Helium-3 is not consumed as fuel and instead circulates in a sealed loop, much as refrigerant in a kitchen fridge is charged once when the appliance is built. Demand therefore follows new dilution refrigerators more closely than total quantum computer sales, since some machines need no helium-3 while others may use several. Current industry forecasts point to substantial growth in the cryogenic infrastructure market, but if this translates into more helium-3-dependent installations, demand for the isotope could rise faster than supply.

How Limited Supply and Competing Users Could Constrain New Installations

Helium-3 production rests on the radioactive decay of tritium, a hydrogen isotope with a half-life of 12.3 years. The US Department of Energy recovers helium-3 at South Carolina’s Savannah River Site from tritium associated with the nuclear weapons programme, with supply historically shrinking as warhead stockpiles were reduced after the Cold War. Canada’s Laurentis Energy Partners extracts the gas from reactor-derived tritium stored at the Darlington nuclear station in Ontario. Russia remains another major producer, but Western access has become politically constrained, with the EU banning imports of Russian helium-3 in 2024. For now, helium-3 output therefore hinges on nuclear defence and power operations rather than quantum computing demand itself, making efforts to establish dedicated new sources, including lunar extraction, particularly significant.

Translating litres into quantum systems makes the supply ceiling tangible. McKinsey’s 2026 Quantum Technology Monitor projects quantum computing revenue rising from just over US$1 billion in 2025 to as much as US$4.4 billion by 2028, and between US$43 billion and US$71 billion by 2035. At the 40-litre helium-3 charge Bluefors specifies for its large XLD1000 refrigerator, a full year of global output would fill approximately 1,000 such systems, assuming every litre went to quantum. Bluefors’ annual purchase ceiling under its Interlune agreement would be equivalent to another 250 systems. These are illustrative comparisons rather than deployment forecasts, since quantum architectures have different cooling requirements and helium-3 can be recycled. Nevertheless, if the expansion of quantum computing drives similarly rapid growth in new millikelvin refrigeration capacity, demand could place increasing pressure on a supply chain that has historically served much smaller markets.

Why New Supply May Arrive from Earth Before The Moon

Unshielded by a magnetic field or atmosphere, the Moon’s surface has absorbed helium-3 from the solar wind for billions of years. Estimates suggest the Moon’s soil could contain around a million tonnes, but turning that into deliveries requires prospecting, excavation, separation, and transport back to Earth. Magna Petra, a start-up developing lunar helium-3, has partnered with NASA to use its mass spectrometer on a planned lunar prospecting mission, and Astrotech (NASDAQ:ASTC), a Nasdaq-listed Texas firm that once built research modules for NASA’s Space Shuttle, has approved a lunar resource strategy that includes helium-3. However, neither plan yet amounts to annual supply.

Lunar ambitions are also already yielding terrestrial technology. Interlune reported in July 2026 that its Cold Capture process produced 99% pure helium-3 from commercially available Grade A helium, and that widespread deployment could roughly triple domestic output. Pulsar Helium, whose Topaz gas in Minnesota contains helium-3 at 12 helium-3 atoms per billion particles of raw gas, independently verified by the USGS and Lawrence Livermore National Laboratory, agreed in October 2026 to evaluate recovery technology with Lunar Helium-3 Mining. Both terrestrial approaches remain early, and the commercial prospects of each will be judged by recovered litres delivered at the right purity and cost.

How Cooling Architecture And Engineering Could Reshape Helium-3 Demand

Future demand also depends on how much gas each unit of computing capacity requires, and that figure is changing. Maybell Quantum’s ColdCloud architecture centralises cooling at 4-kelvin and distributes it to independent millikelvin nodes. The Colorado company claims up to 80% less helium-3 per qubit than an equivalent array of conventional dilution refrigerators, with first commercial deployment planned for 2027. In August 2026, IBM successfully connected two cryogenic modules into one shared environment below 15 millikelvin (more than 180 times closer than deep space). This is a key step towards linking hundreds of chips for Starling, its fault-tolerant system due in 2029. Modularity lets a system’s cooling expand piece by piece as its processors grow, so even as cooling becomes more efficient, total helium-3 demand could keep climbing with every module added.

Other approaches remove helium-3 from parts of the cooling chain altogether. Sydney’s Diraq reported in Nature in 2024 that its silicon spin qubits ran above 1 kelvin with fidelities in the range required for error correction, a temperature achievable with helium-4 cooling. In a February 2026 paper in Review of Scientific Instruments, Germany’s Kiutra and the Netherlands’ Delft Circuits kept a system at 30 millikelvin for 24 hours without helium-3 using continuous magnetic refrigeration. Continuous magnetic refrigeration cools specialist materials by slowly switching off a magnetic field around them, with several units taking turns so the cooling never stops. Both results must still prove themselves in larger systems, yet they show that helium-3 dependence is an engineering choice, and such choices may change once an input becomes expensive enough.

The market is also separating by temperature and application. In September 2026, Canada-based Xanadu and Bluefors announced plans for a prototype module to cool the single-photon detectors in Xanadu’s photonic quantum computers near 2 kelvin, potentially eliminating the need for industrial-scale cryoplants. The technology remains a development project, but illustrates how cryogenic suppliers could benefit from quantum computing growth without a corresponding increase in helium-3 demand.

Where the Bottleneck Would Likely Bite First

Helium-3 becomes a significant constraint only when demand for qualified gas exceeds what buyers can obtain through production, inventories, and reuse within their deployment schedules. A shortage would most likely slow the creation of new capacity while leaving the existing quantum computing fleet running, favouring operators and suppliers that already hold gas over entrants still assembling infrastructure. The evidence will surface across various measurable factors, from annual deliveries and buyer lead times to gas inventory per installation and alternative cooling proven under realistic loads.

For investors and policymakers, the question is whether helium-3 from tritium today, and from helium plants, gas fields, and perhaps the Moon in future, can arrive quickly enough, alongside leaner cooling, to stay ahead of the installations that quantum computing forecasts imply. The pace at which quantum computing scales may depend as much on that race as on the qubits themselves.

5