The Cryogenic Infrastructure That Makes (or Breaks) a Quantum Computer

Introduction

A superconducting transmon qubit operates at 10-20 millikelvin. Deep space sits at 2.7 kelvin. The mixing chamber of a dilution refrigerator is roughly 150 times colder than the cosmic microwave background, and maintaining that temperature continuously while pumping microwave signals through hundreds of cables into a chip that must remain coherent for tens of microseconds is the central engineering challenge of superconducting quantum computing.

The cryogenic infrastructure is also the part of the build that surprises first-time buyers the most. A dilution refrigerator is not a piece of lab equipment you rack-mount and forget. It is an industrial installation that weighs three-quarters of a metric ton, takes months to procure, costs more than many of the QPUs it will host, runs on a working fluid derived from nuclear weapons stockpile decay that costs $2,500 per liter, and will shut down your entire quantum operation for a week if the power goes out.

This article covers the three cryogenic constraints that dominate the cost, timeline, and scaling ceiling for any quantum computer that operates below 4 K: the dilution refrigerator market, the helium-3 supply chain, and the I/O wiring wall. These topics apply directly to superconducting and silicon-spin builds, partially to photonic builds (for the SNSPD detector subsystem), and not at all to neutral-atom or most trapped-ion builds. The fact that two of the five major qubit modalities avoid cryogenics entirely is itself a strategic consideration for procurement decisions.

The dilution refrigerator market

A dilution refrigerator is the only technology that provides continuous cooling to millikelvin temperatures. It works by exploiting a quantum thermodynamic property of helium-3/helium-4 mixtures: below approximately 870 mK, the mixture spontaneously separates into two phases, a concentrated phase (mostly helium-3) and a dilute phase (mostly helium-4 with approximately 6.6% helium-3). Helium-3 atoms crossing the phase boundary from the concentrated phase into the dilute phase absorb energy from the surrounding environment, cooling it. By continuously circulating helium-3 through this phase boundary (using pumps and heat exchangers), the system maintains continuous cooling at millikelvin temperatures indefinitely. The pulse-tube cryocooler handles the initial cooling from 300 K to approximately 3 K; the dilution circuit takes over from there to below 10 mK.

The distinction between helium-3 and helium-4 matters because they have completely different supply chains. Helium-4 (the common isotope, used in MRI cooling, balloon inflation, and as a cryogenic fluid) is extracted from natural gas wells and costs $5-10 per liter. Helium-3 (the rare isotope, used as the working fluid in dilution refrigerators and as a neutron detector material) derives from tritium decay in nuclear weapons stockpiles and costs $1,900-$2,600 per liter. The 2026 helium supply disruptions from the Qatar conflict primarily affect helium-4. The helium-3 bottleneck is a separate, structural problem driven by the finite rate of tritium decay in a shrinking global weapons stockpile.

The dilution refrigerator market is consolidated around a small number of specialist vendors.

Bluefors (Helsinki, Finland) is the market leader with over 1,800 systems shipped globally. The product line spans five tiers:

The LD450sl (previously LD400sl, renamed March 2026) is the research workhorse. Base temperature below 10 mK. Over 450 µW cooling power at 100 mK. Preconfigured systems ship in approximately 4 months, the fastest delivery in the market. Price range: $700K-$1M. The “sl” suffix denotes side-loading ports for cryogenic wiring access. Best for systems up to approximately 30 qubits.

The XLD1000sl targets 30-400 qubit builds. Over 1,000 µW cooling power at 100 mK. Dual pulse-tube cryocoolers (Cryomech PT425-RM). Pre-installed Cri/oFlex integration through the Bluefors/Delft Circuits partnership: each side-loader supports up to 256 high-frequency lines in configurations of 64, 128, or 256 channels, allowing up to 1,536 lines in an XLDsl system. Holds approximately 40 liters of helium-3 (~$100,000 at current prices). Lead time: 6-12 months. Price range: $1.5M-$2.5M.

The KIDE is the industrial platform for 400-1,000+ qubits. Over 3,000 µW cooling power at 100 mK from three independent cooling units. Hexagonal self-standing chamber approximately 3 m tall by 2.5 m diameter, with full-height access doors. 1.6 m² MXC flange area. Up to 500 kg payload. 24 side-loading ports plus top-loading, supporting over 4,000 high-density wiring RF lines. Nine pulse tubes. Long-life cold trap enables continuous operation up to 3 years between full warm-ups. First 18 systems shipped to the AIST G-QuAT center in Tsukuba, Japan (May 2025). Used inside IBM Quantum System Two. Price range: $5M-$10M. Weight: up to ~7,000 kg. Lead time: 6-12 months.

The XLDHe High Power addresses silicon-spin and SNSPD workloads operating at approximately 1 K rather than 10-20 mK. Provides 200-700 mW of cooling power at 1-1.2 K using helium-4 only. No helium-3 required. This eliminates the most strategically constrained input in the quantum supply chain for modalities that can operate at higher temperatures. See the silicon-spin build guide for the implications.

The Modular Cryogenic Platform, announced March 2026, is the next-generation architecture for scaling beyond KIDE. Self-supported expandable vacuum chambers where individual modules can be interconnected to create a unified, continuous payload space. Cooling and wiring are functionally decoupled: operators can upgrade cooling capacity or reconfigure wiring independently without redesigning the entire cryogenic environment. Each module supports payloads up to 800 kg with up to 36 side-loading wiring ports. Low-height form factor designed for HPC data center integration. First multi-module delivery slated for late 2026. This platform is designed for the era of thousands to hundreds of thousands of qubits.

Note on Bluefors nomenclature: In March 2026, Bluefors updated its product naming and published unified guaranteed cooling power specifications. The LD250 became LD350; LD400/LD400sl became LD450/LD450sl; XLD400sl became XLD450sl. The updated guaranteed figures replace the prior two-tier “guaranteed” and “expected” system. If you are comparing quotes from before and after March 2026, verify you are using the current naming.

Maybell Quantum (Denver, Colorado) builds the Big Fridge. Base temperature below 10 mK. Over 1,000 µW guaranteed cooling power at 100 mK. Approximately 130-liter sample volume. Over 3,400 cm² MXC plate area. Integrated Minus K Technology negative-stiffness vibration isolators rated below 0.5 Hz, decoupling pulse-tube cryocooler vibration from the dilution unit. Integrated LF CryoTrace flexible ribbon technology (MIT Lincoln Laboratory license) for low-frequency services. Design philosophy eliminates scroll pumps, acid-flux solder, rubber gaskets, and KF flanges in favor of welded and metal-to-metal joints, reducing maintenance touch points. Lead time: approximately 6 months. Price range: $1.5M-$3M. Best for: US sovereign supply chain requirements, builds where integrated vibration isolation is critical, 30-200 qubit range. Maybell also signed a multi-year He-3 agreement with Interlune.

Oxford Instruments NanoScience (Abingdon, UK) builds the Proteox MX/LX. Modular dilution refrigerator with rapid sample exchange. LX variant provides over 450 µW class cooling power at 100 mK. Oxford reports a 50% lead-time reduction from prior baselines via a factory-reconfiguration program (informally 5-8 months). The TeslatronPT wet cryostat with pulse tube has also seen roughly two-thirds lead-time reduction. Best for: academic groups needing frequent QPU swaps, multi-project research environments, UK supply chain.

Leiden Cryogenics (Netherlands) builds the CF-CS series. Heritage academic supplier with custom geometries and deep expertise in dilution refrigerator design. Serves research groups requiring non-standard configurations.

Kiutra (Munich, Germany) is developing the only commercial He-3-free alternative that can reach millikelvin temperatures for superconducting qubits. Their continuous adiabatic demagnetization refrigeration (cADR) technology uses the magnetocaloric effect of paramagnetic refrigerants to cool without helium-3 or any cryogenic liquids. A February 2026 paper in Review of Scientific Instruments demonstrated continuous sub-30 mK operation with integrated high-density RF wiring for a five-qubit superconducting processor. Kiutra raised €13 million (bringing total funding to over €30M from Intel Ignite, Trumpf Venture, and EIC Accelerator) specifically to expand modular platforms tailored for quantum chips. Their EIC-funded LEMON project is developing large-scale helium-3-free cooling. The limitation: cADR provides pulsed cooling with limited duty cycle compared to the continuous cooling of dilution refrigeration. The February 2026 demonstration achieved continuous operation, which is a significant step, but the cooling power at millikelvin temperatures is currently lower than comparable dilution refrigerators. For research systems and small qubit counts, Kiutra’s cADR is already viable. For industrial-scale deployments requiring continuous high-power cooling at 10-20 mK, dilution refrigeration remains the standard. Track Kiutra’s LEMON project for scaling updates.

Helium-3: the $2,500-per-liter strategic bottleneck

Helium-3 is the working fluid in every dilution refrigerator. It does not occur naturally on Earth in commercial quantities. Virtually all terrestrial supply derives from the radioactive decay of tritium (half-life 12.3 years) in nuclear weapons stockpiles, primarily in the United States. Russian supply has been effectively excluded from Western markets since 2022. Global terrestrial production is estimated at 22,000-30,000 liters per year (sources vary; the Lowy Institute’s 2025 synthesis cites The Edelgas Group data). Demand from quantum computing, medical imaging (He-3 lung MRI), neutron detection (homeland security and nuclear safeguards), and experimental fusion research is estimated at 40,000-60,000 liters per year and rising.

Market pricing ranges from $1,900 to $2,600 per liter for bulk supply, with purification costs exceeding $10,000 per liter. A Bluefors LD450sl holds approximately 5-12 liters of helium-3. An XLD1000sl holds approximately 40 liters. KIDE-class systems require substantially more. At $2,500 per liter, the helium-3 charge alone in an XLD represents a $100,000 investment in a fluid that must be recovered (not vented) after every servicing event.

The structural deficit is roughly 20,000-40,000 liters per year: demand exceeds terrestrial supply and the gap is growing as quantum computing scales. The 2026 helium supply crisis (Qatar supply disruptions from the US-Iran conflict have affected helium-4 infrastructure, drawing attention to broader helium geopolitics) has heightened awareness, though it is important to distinguish helium-3 supply chains (nuclear weapons stockpile decay) from helium-4 supply chains (natural gas extraction byproduct). The two isotopes are distinct markets with different drivers.

The lunar hedge

Three forward-purchase agreements represent the industry’s strategic hedge against terrestrial helium-3 scarcity:

Bluefors signed an agreement with Interlune on 16 September 2025 to purchase up to 10,000 liters per year of lunar-regolith-extracted helium-3 for 2028-2037 deliveries. Reported contract value: ~$300 million over the term. Maybell Quantum signed a separate multi-year He-3 agreement with Interlune. The U.S. Department of Energy also contracted with Interlune for He-3 supply through its Isotope Program.

Interlune’s extraction technology centers on an advanced mass spectrometer lunar harvester. The Prospect Moon payload (NASA SBIR Phase III, $6.9M, awarded May 2025) targets a 2027 extraction demonstration on the Astrolab FLIP rover launched on Astrobotic’s Griffin 1 lander, with an end-to-end Earth-return mission slated for 2028. Interlune CEO Rob Meyerson has indicated that commercial-scale lunar extraction will be in the early 2030s.

These timelines are technologically aggressive. No helium-3 from non-terrestrial sources should be assumed available before 2029 at the earliest, and commercial volumes before the early 2030s. Integrators should treat the Interlune agreements as long-term strategic hedges, not as near-term procurement dependencies.

Practical guidance for helium-3 management

Any organization operating one or more dilution refrigerators should: negotiate helium-3 supply as part of the cryostat purchase contract (vendors have allocation arrangements that individual buyers do not). Install closed-loop gas handling and helium recovery infrastructure from day one (Quantum Technology Corp HR3/HRHP series is the off-the-shelf choice). Maintain at least one warm-spare helium-3 charge on site for contingency replacement. Never vent helium-3 to atmosphere during servicing. Budget for the recovery and purification cycle after every warm-up event. If the He-3 spot price doubles within a year, forward-buy a multi-charge reserve and accelerate recovery system commissioning. Track the Interlune and Magna Petra lunar timelines for planning purposes, but do not treat them as procurement dependencies.

The I/O wiring wall

Every superconducting qubit requires multiple RF and DC control and readout lines connecting room-temperature electronics to the millikelvin processor. A 20-qubit system with tunable couplers can require 80-120 distinct cryogenic signal paths. A 100-qubit system demands 300-500. A 1,000-qubit system requires 3,000-5,000.

Conventional coaxial cables (stainless steel and NbTi semi-rigid, 2.19 mm outer diameter per line) top out at approximately 168 channels per cryostat loader. The physics is straightforward: coax is too thick, too thermally conductive (each cable segment conducts heat from warmer to colder stages), and too rigid to scale. A naive coaxial wiring installation for a 100-qubit system can exhaust the cryostat’s 100 mK cooling power budget on attenuator heat loads alone, leaving no thermal margin for the QPU.

This is the I/O wiring wall, and it is the binding constraint on qubit scaling for superconducting systems, independent of QPU quality. A QPU vendor can fabricate a 1,000-qubit chip, but if the cryostat cannot support 3,000-5,000 signal paths without exceeding its cooling budget or physical space, the chip cannot be used.

Cri/oFlex: the commercial answer

Delft Circuits’ Cri/oFlex is the leading solution. NbTi superconducting stripline on polyimide substrate, 0.3 mm thick (compared to 2.19 mm for coax), 8 channels per flex ribbon, 50 Ω controlled impedance, 1.5 dB/m insertion loss at 6 GHz, less than 10 mΩ contact resistance. Attenuators, low-pass filters, and IR filters are integrated directly into the flex structure, eliminating dozens of manual splice points per line. Thermal anchoring is achieved by clamping the flex ribbon at each cryostat stage.

Current density: 256 channels per 108 mm side-loading port (Bluefors XLDsl integration). Published roadmap (September 2025): 1,024 channels per loader by 2027, 4,096 by 2029. This roadmap defines the superconducting qubit scaling ceiling more directly than any QPU roadmap. If Delft Circuits delivers on the 4,096-channel target, a KIDE-class cryostat with 24 side-loading ports could theoretically support over 98,000 RF lines, enough for approximately 20,000-30,000 qubits depending on the control architecture.

Bluefors launched its own High-Density Flex Wiring (FPC) platform in March 2025, offering up to 240 channels per side-loading port. This competes with Delft Circuits at the lower end but is designed for tight integration with Bluefors’ own cryostat side-loading architecture. The two are complementary in practice: Bluefors pre-installs Cri/oFlex in its XLDsl-series fridges through their strategic partnership.

MIT Lincoln Laboratory’s LF CryoTrace flexible ribbon technology is the US-pedigree alternative, licensed to Maybell Quantum. It handles low-frequency services (thermometry, heaters, DC gate voltages) rather than high-frequency qubit control, and is complementary to Cri/oFlex rather than competitive.

FormFactor partnered with Delft Circuits in March 2025 for cryogenic probing and testing integration, providing up to 160 channels per port (1,920 total) in FormFactor’s cryogenic test systems.

The integration imperative

An integrator starting a new superconducting build in 2026 should not accept a coax-only wiring design for any system above 20 qubits. The upgrade path from coax to flex requires a full warm-up, complete rewire, and cool-down cycle: 2-4 weeks of downtime plus the risk of disrupting a working system. Specifying Cri/oFlex or equivalent from day one avoids a costly future retrofit. The cost difference between a coax wiring tree and a Cri/oFlex installation is small relative to the total system cost, and the scaling headroom it provides is essential for any QPU upgrade path.

Cooling power budgets

Every component in the cryostat contributes to the thermal load at each temperature stage. The integrator must verify that the total heat load at each stage stays within the cryostat’s cooling capacity, with margin for future QPU and wiring expansion.

A rule-of-thumb heat-load budget for a 100-200 qubit superconducting system: over 1 W at 50 K, over 700 mW at 4 K, over 5 mW at 100 mK (cold plate), over 500 µW at MXC (20 mK). The signal chain described in the superconducting build guide details where this heat comes from: each coaxial cable segment conducts heat along its length, each attenuator dissipates power as heat at its mounting stage, each amplifier (HEMT, TWPA) generates heat at its operating point. Running a 100-qubit setup with conventional coaxial wiring can exhaust the 100 mK cooling power budget on attenuators alone. This is why the I/O wall is a thermal problem as much as a physical-space problem.

For integrators, the practical message: request the detailed cooling power curve from the cryostat vendor (cooling power vs. temperature at each stage) and run a heat-load calculation before finalizing the signal chain design. If the total heat load at any stage exceeds 80% of the available cooling power, you have no margin for QPU upgrades or additional instrumentation. Either redesign the signal chain (reduce attenuation, move components to warmer stages) or specify a cryostat with higher cooling power.

Cool-down, warm-up, and maintenance timelines

Cool-down. An empty Bluefors LD system reaches base temperature in under 24 hours. An LD/XLD with full payload (QPU, hundreds of RF/DC lines, magnetic shields) takes 3-7 days. KIDE takes longer due to its larger thermal mass. A fully loaded XLD1000sl with a QPU and Cri/oFlex wiring should budget 5-7 days from room temperature to stable base temperature.

Warm-up. Controlled warm-up to room temperature for service takes 2-4 days. The warm-up must be controlled (not vented to atmosphere) to avoid mechanical stress on connectors, seals, and the QPU package from thermal shock.

Maintenance cadence. Pulse-tube cold-head service every 18-24 months (rotary valve and motor replacement). Compressor adsorber replacement annually. Helium-3 mixture composition check annually. Gas-handling system filter changes per vendor schedule. Quarterly leak checks. Annual full health audit. Plan at least one planned warm-up and service window per year: budget 5-10 days of total downtime (warm-up, service, cool-down, recalibration).

KIDE’s long-life cold trap enables continuous operation up to 3 years between full warm-ups, a significant operational advantage for production systems where every warm-up costs a week of downtime plus QPU recalibration.

The unplanned warm-up. The costliest operational event in superconducting quantum computing. A power outage that takes the pulse-tube cryocooler offline, a compressor failure, or a cooling-water interruption will warm the cryostat. If the MXC exceeds 1 K, the recovery process takes a minimum of 5-10 days: recover the helium-3 gas (do not vent), diagnose the root cause, repair, cool back down, run the full QPU recalibration sequence. At $10K-$50K per day in personnel idle time and lost user access, a single unplanned warm-up event is a five- to six-figure incident. This is why the facility preparation guide specifies online double-conversion UPS, redundant chilled-water loops, and automatic shutdown sequencing as non-negotiable infrastructure.

Spare parts that prevent downtime

For a single-cryostat production system, the playbook recommends maintaining on site: one spare pulse-tube cold head (rotary valve and motor kit, because vendor swap can take days and the service interval is 18-24 months), two spare compressor adsorber cartridges (annual replacement), one spare helium-3 charge in a sealed cylinder (in case of mixture contamination or leak during servicing), a stock of O-rings and vacuum seals (viton and indium, used during every warm-up service), spare SMA/SMP connectors (these are consumables damaged by torque wrenching), a spare HEMT amplifier (Low Noise Factory LNC4_8C or equivalent, because lead times for new HEMTs can stretch to months), and a spare TWPA (Silent Waves Zephyr or equivalent, same lead-time concern). Estimated spare-parts inventory cost: $80K-$150K. This is insurance against downtime that costs $10K-$50K per day.

The cryo-CMOS horizon

The constraints above (helium-3 supply, I/O wiring density, cooling power budgets) apply to the current architecture where room-temperature control electronics connect to millikelvin qubits through long signal chains. The principal architectural shift on the 2028 horizon is cryo-CMOS co-integration: placing control electronics at cryogenic temperatures adjacent to or on the same chip as the qubits.

Intel’s Horse Ridge II (22 nm FinFET SoC operating at 3 K, 128 qubits addressable through 4 RF channels via frequency-division multiplexing) and Pando Tree (millikelvin-stage controller) represent the most advanced programs. SEEQC’s single-flux-quantum digital logic at 4 K demonstrated over 99.9% gate fidelity (December 2025). SemiQon’s cryo-CMOS transistors achieve record 0.3 mV/dec sub-threshold swing at 420 mK. QuTech demonstrated cryo-CMOS control for diamond NV centers with no observable qubit-fidelity degradation (February 2026). Diraq and Emergence Quantum are co-integrating millikelvin CMOS control with silicon-spin qubits.

Cryo-CMOS will not replace room-temperature racks before approximately 2028, but it is the only credible engineering path to controlling more than 10,000 qubits without the I/O wiring wall collapsing. When cryo-CMOS reaches production maturity, the cryostat’s role shifts from “bridge between room temperature and millikelvin” to “thermal enclosure for a co-integrated quantum-classical chip.” The signal chain simplifies. The wiring density problem dissolves. The helium-3 requirement remains (for transmon qubits), but the cooling power budget changes because the cryo-CMOS electronics generate heat at the same stage where they operate.

What this means for procurement

The cryostat selection and helium-3 management strategy should be the first decisions in any superconducting or silicon-spin quantum computer procurement, made before QPU selection, before control electronics selection, and before facility design is finalized. The cryostat has the longest lead time (4-12 months), the highest capital cost ($700K-$10M), and the longest operational lifetime (a decade or more). It is the one component you will not upgrade on a 2-3 year cycle.

Build for the next system, not the current one. If your roadmap targets 100+ qubits within five years, do not buy an LD450sl to save money today. The cost of operating an undersized cryostat and then replacing it will exceed the initial differential.

Specify flex cabling from day one. A coax-only design is a future retrofit cost.

Negotiate helium-3 supply with the cryostat purchase. Install recovery infrastructure before the first cool-down.

For organizations where helium-3 supply risk is unacceptable (defense customers, sovereign quantum programs with supply chain constraints), track the silicon-spin modality (Bluefors XLDHe, He-4 only at 1 K), Kiutra’s cADR program (He-3-free millikelvin cooling), and neutral-atom systems (no cryogenics at all) as strategic alternatives to dilution-refrigerator-dependent modalities.

For the signal chain that runs inside the cryostat, see the superconducting build guide. For the facility that houses the cryostat, see the facility preparation guide. For the total cost of ownership including cryogenic operational costs, see the cost and procurement article.