Preparing a Facility for a Quantum Computer

Introduction

In September 2025, IQM engineers surveyed three potential locations inside the Leibniz Supercomputing Centre in Munich before selecting a spot for Germany’s first hybrid HPC-quantum deployment. The survey lasted over 25 hours. They measured floor vibrations, magnetic fields, and sound levels across a full diurnal cycle, discovering that passing trams and even music playing in an adjacent room introduced measurable noise into the frequency bands that matter for superconducting qubit coherence. The data from that survey, published in arXiv:2509.12949, confirmed something that every quantum hardware builder already knew but that most data center operators have never confronted: a quantum computer has facility requirements that no conventional server room is designed to meet.

This does not mean quantum computers cannot go into existing buildings. The LRZ team proved they can. But it means the gap between what a standard data center provides and what a quantum computer demands must be measured, understood, and closed before hardware arrives. The cryostat, the QPU, and the control electronics can all be procured in parallel with facility remediation. The facility work should start first, because discovering a structural floor-loading deficiency or an EMI problem after the cryostat is delivered is a schedule disaster.

This article covers every parameter that matters, organized by what fails most expensively if you get it wrong. For the broader integration context, see the capstone article in this series. For the specific integration sequences by qubit modality, see the superconducting, trapped-ion, neutral-atom, and silicon-spin build guides.

The site survey: what you measure and why it takes 25 hours

A proper site survey for a superconducting quantum computer installation runs at least 25 hours of continuous measurement. The duration matters because many interference sources are periodic: HVAC cycling, elevator traffic, commuter rail schedules, building cleaning crews operating heavy equipment at night, and diurnal temperature swings that shift structural resonances. A four-hour survey on a quiet Saturday afternoon will miss the Monday-morning freight elevator that shakes the floor at a frequency your transmon qubits cannot tolerate.

The survey measures five categories of environmental parameters, each with defined pass/fail criteria.

Vibration. Accelerometer data across the full frequency spectrum, evaluated against the VC-A vibration criterion (or stricter). VC-A is the standard for microsurgery and electron microscopy; quantum computing installations at the state of the art go further. The critical frequency range is below 10 Hz, where building-borne vibrations from traffic, foot traffic, and HVAC compressors dominate. A separate concern is the pulse-tube cryocooler’s own vibration at approximately 1 Hz, generated by helium gas flow in the compressor. Maybell Quantum’s Big Fridge addresses this with integrated Minus K Technology negative-stiffness isolators rated below 0.5 Hz, decoupling the pulse-tube oscillation from the dilution unit. For other cryostats, the vibration-isolation pad or slab beneath the fridge must handle both building-borne and cryocooler-generated vibration.

Electromagnetic interference (EMI). Spectrum analysis across the 4–8 GHz qubit operating band, plus DC and AC magnetic field mapping. The OCP white paper specifies DC magnetic fields below 100 µT and AC magnetic fields below 1 µT in the vicinity of the cryostat. Standard overhead fluorescent lighting generates unacceptable EMI at these thresholds. So do certain UPS topologies, variable-frequency motor drives in HVAC systems, and cellular base stations within 100 meters.

Power quality. Total harmonic distortion (THD), voltage sags and swells, 3-phase balance, and available amperage at the proposed panel location. The target is below 1% THD on every outlet feeding the control racks and cryostat support systems.

Structural. Floor loading capacity at the proposed cryostat location, measured in kg/m² and verified against the specific cryostat model being installed.

Thermal and humidity. Ambient temperature stability (target: 20–24°C ± 2°C), relative humidity (30–60% non-condensing), and the availability and routing of chilled-water supply at the required temperature and flow rate.

Applied Quantum provides turnkey site surveys as the first phase of any integration engagement. The deliverable is a facility specification document identifying every parameter that meets, exceeds, or fails the required threshold, with a remediation plan and cost estimate for any gaps.

Floor loading: the weight problem nobody expects

A fully loaded Bluefors LD or XLD-class dilution refrigerator weighs approximately 750 kg. A KIDE-class platform can reach roughly 7,000 kg. This weight is concentrated on a small footprint, not distributed across a wide area like a standard server rack.

Standard raised flooring in conventional data center colocation facilities is generally insufficient. Raised-floor tiles in a typical data center are rated for 300–500 kg/m² distributed load. A 750 kg cryostat on a 1 m² footprint exceeds that capacity, and the point-loading from the cryostat legs concentrates the force further. The OCP white paper and Applied Quantum’s playbook both specify a minimum floor loading capacity of 1,000 kg/m².

The solution in most installations is either steel reinforcement pedestals (for raised-floor environments) or direct slab-on-grade mounting (for purpose-built quantum labs). The IQM/LRZ team reports that their installation required verifying structural capacity before positioning the system. The Q-PAC deployment at the Quantum Commons campus in Denver benefited from a purpose-built facility. For retrofit installations in existing buildings, a structural engineer must sign off on the floor loading at the specific proposed location before the cryostat order is placed.

The delivery path also matters. The cryostat arrives in wooden crates weighing approximately 750 kg. Corridors, doorways, and freight elevators along the delivery path must accommodate these dimensions and weights. An XLD or KIDE will not fit through a standard office doorway.

Vibration isolation: 100 meters from the nearest elevator

Two vibration sources must be managed independently: external building-borne vibration and internal cryocooler vibration.

External vibration. Low-frequency vibration below 10 Hz from traffic, elevators, HVAC plant, foot traffic, and nearby rail or truck routes couples through the building structure to the cryostat and ultimately to the QPU at the mixing chamber. The IQM/LRZ team reports that vibration below 400 µm/s was achieved with careful site selection within the existing facility. The playbook specifies a minimum exclusion zone of 100 meters from heavy infrastructure sources: elevator shafts, train lines, freight loading docks, heavy HVAC compressors, and cellular base stations.

The reference floor plan positions the cryostat on a vibration-isolated slab or plinth, built to VC-A criterion or stricter, preferably slab-on-grade (directly on the building’s structural slab, not on raised flooring or upper-level suspended floors). A service clearance zone of at least 1.2 meters surrounds the cryostat on all sides, allowing access for can removal, side-loading port access, and maintenance.

Internal cryocooler vibration. Every pulse-tube cryocooler generates mechanical vibration at approximately 1 Hz from the reciprocating helium gas flow in the cold head. In most dilution refrigerators (Bluefors, Oxford), the pulse tube is mechanically coupled to the cryostat body. Maybell Quantum’s Big Fridge integrates Minus K Technology negative-stiffness vibration isolators that decouple the pulse-tube motion from the dilution unit, rated below 0.5 Hz. For other cryostats, the vibration-isolation pad beneath the system must absorb both building-borne and cryocooler-generated vibration.

Remote-motor pulse-tube configurations (Cryomech PT415-RM, PT425-RM, PT310-RM) move the motor away from the cryostat body, reducing the vibration transmitted to the cold head. This is standard in most research-grade installations.

EMI shielding: why your fluorescent lights are a problem

Superconducting transmon qubits operate at microwave frequencies between 4 and 8 GHz. Any electromagnetic interference in this band couples directly to the qubits and degrades coherence. The required shielding is more stringent than anything a conventional data center provides.

DC magnetic fields must be below 100 µT at the cryostat location. For reference, the Earth’s magnetic field is approximately 25–65 µT, so the ambient DC field is near the threshold in many locations before any building infrastructure is considered. Steel structural elements, reinforced concrete, and nearby electrical equipment all create local DC field variations.

Inside the cryostat, the QPU is protected by layered magnetic shielding: an outer mu-metal can (high-permeability nickel-iron alloy, Cryoperm from Vacuumschmelze or A4K from Amuneal) for DC field rejection, and an inner superconducting can (niobium or lead) for high-frequency rejection. These shields are installed as part of the signal chain integration. But the room-level EMI environment still matters because the drive and readout lines are exposed between the top plate of the cryostat and the control racks.

AC magnetic fields must be below 1 µT. Standard overhead fluorescent lighting generates broadband electromagnetic noise that fails this threshold. The fix is simple: replace fluorescent fixtures within at least 2 meters of the cryostat (preferably the entire quantum lab) with low-noise LED arrays driven by DC power supplies. Variable-speed motor drives in HVAC systems are another common AC EMI source; route HVAC ducting to avoid passing directly over or adjacent to the cryostat.

Room-level shielding. The reference floor plan shows the entire quantum lab enclosed in a Faraday enclosure for EMI control. This is a conductive mesh or plate enclosure (copper or aluminum) surrounding the room, grounded to the building’s equipotential bonding system. All cable and pipe penetrations through the enclosure walls must be sealed and EMI/vibration isolated. Star-topology grounding from a single reference point prevents ground loops. Isolation transformers on the power feed reject common-mode noise. Fiber-optic isolation on all digital interfaces touching the cryostat eliminates conducted EMI paths.

For installations where a full Faraday room is cost-prohibitive, a localized RF-shielded enclosure around the cryostat and immediate control racks provides a practical alternative. The IQM/LRZ installation operated successfully without a full Faraday room, relying on careful site selection and localized shielding.

Power: 3-phase, UPS, and why grid loss costs you a week

A superconducting quantum computer’s power demand is modest by data center standards. The IQM/LRZ paper reports peak power consumption of approximately 30 kW during cool-down (substantially less than a single classical Cray EX4000 cabinet at ~140 kW), dropping to 10–15 kW steady-state depending on the number of control instruments. But the quality and reliability of that power are far more critical than the quantity.

Power specification. Dedicated 3-phase 63 A panel, conditioned to below 1% total harmonic distortion on every outlet feeding the control racks and cryostat support. Online double-conversion UPS sized for at least 20 minutes of full load. The UPS does not need to power the system indefinitely. It needs to ride out grid sags and transients and provide enough time for a controlled warm shutdown that protects the helium-3 mixture. If the cryostat has an automated shutdown sequence (Bluefors Gen-2 control software supports this), 20 minutes is sufficient. If manual intervention is required, 30 minutes provides more margin.

Why this matters so much. A grid loss that takes the pulse-tube cryocooler offline will warm the cryostat. If the mixing chamber exceeds 1 K (which happens within hours without cooling), recovery requires a minimum of 5–10 days: controlled warm-up to room temperature, helium-3 gas recovery (the mixture must not be vented), diagnosis of what caused the power loss, repair, cool-down (3–7 days), and full recalibration. At $10K–$50K per day in personnel idle time and lost user access, a single unplanned warm event from a preventable power interruption is a five- to six-figure incident.

Redundant power feeds (A+B) from separate utility circuits reduce the probability of a total outage. A backup generator with automatic transfer switch provides an additional layer for facilities where grid reliability is marginal.

Cooling water: colder than your HPC liquid-cooling loop

Pulse-tube compressors (Cryomech PT415-RM, PT425-RM, PT310-RM with remote motors) require chilled water at 15–25°C, with flow rates of 10–30 L/min at 4–6 bar pressure. This is substantially colder than the approximately 45°C warm-water loops used in modern energy-efficient HPC liquid-cooling systems (such as the hot-water cooling at LRZ’s SuperMUC-NG). A quantum computer cannot share a warm-water HPC cooling loop.

A dedicated secondary chilling loop is required. The reference floor plan shows the chiller and cryogenic support system adjacent to but outside the shielded quantum lab, with insulated, vibration-isolated chilled-water supply and return lines penetrating the enclosure wall. The vibration isolation on these lines is important: chiller compressors generate mechanical vibration that can propagate through rigid piping directly to the cryostat.

For multi-cryostat installations, the chilled-water plant must be sized for the total heat rejection of all pulse-tube compressors operating simultaneously, plus margin for redundancy.

Helium storage and recovery: the $2,500-per-liter insurance policy

Every dilution refrigerator contains a charge of helium-3 and helium-4 gases. Helium-3, the working fluid that enables cooling below 300 mK, costs $1,900–$2,600 per liter at current market prices, with purification costs exceeding $10,000/L. A Bluefors XLD1000sl holds approximately 40 liters, roughly $100,000 of working fluid.

This gas must be recovered after every servicing event that requires warming the cryostat. If the mixture is vented to the atmosphere, the cost of replacement is significant and the lead time for procuring new helium-3 can stretch to months. Closed-loop gas handling is mandatory, not optional.

The reference floor plan shows a dedicated helium storage enclosure, physically separated from the quantum lab. Key design requirements: secured, ventilated cage or room with restricted access (keycard or equivalent). Oxygen sensors at floor level and ceiling level (helium is lighter than air; helium-4 rises, but leak scenarios can displace oxygen at any level in enclosed spaces). Closed-loop gas handling system (Quantum Technology Corp HR3/HRHP series) connecting the cryostat to recovery cylinders. Helium-3 storage cylinders rated for the working pressure of the gas-handling system. At least one warm-spare helium-3 charge stored on site for contingency replacement.

The oxygen displacement risk is the primary safety concern. Both helium-3 and helium-4 are asphyxiation hazards in enclosed spaces. A sudden release from a gas-handling system failure can rapidly displace breathable air. Floor-level and ceiling-level oxygen monitors with audible and visual alarms are mandatory in both the quantum lab and the helium storage enclosure.

Room layout: the reference floor plan

The reference floor plan [shown above] illustrates the spatial relationships and clearances for a single-cryostat superconducting quantum computer installation. Key dimensions and requirements:

Ceiling height: minimum 3 meters clear. An XLD or KIDE cryostat plus its top-loading wiring tree is tall. Overhead crane or hoist access, centered on the cryostat, is needed for cool-down servicing, wiring tree installation, and QPU swap operations. The crane rated load must accommodate the heaviest component being lifted (typically the vacuum can or radiation shields, up to 100–200 kg for an XLD, substantially more for KIDE).

Cryostat footprint and clearance. An IQM 20-qubit system including cryostat and racks occupies approximately 126 × 453 × 290 cm. Maintain a service clearance zone of at least 1.2 meters around the cryostat on all sides for can removal and side-loading port access. The reference plan shows approximately 2–3 meters between the cryostat and the control electronics racks; cable runs from racks to the cryostat top plate should be kept below 2–3 meters where possible to minimize signal degradation and cable management complexity.

Control electronics racks sit within the shielded quantum lab, connected to the cryostat top plate via RF coaxial cables or Cri/oFlex feed-throughs. Standard 19-inch rack format. The number of racks depends on qubit count and control vendor: a 20-qubit Qblox Cluster configuration fits in 1–2 racks; a 64-qubit system may require 3–4 racks with additional local oscillator sources.

Network patch panel inside the quantum lab connects to the facility’s HPC network via fiber (10/25/100 GbE dedicated VLAN to the HPC). For NVQLink integration, the GPU node may sit in the quantum lab (within the 400 Gb/s Ethernet distance limit) or in an adjacent HPC room. See the HPC integration guide for network design details.

UPS and power conditioning sit adjacent to or within the quantum lab. The reference plan shows them outside the shielded enclosure but within the same room suite, to avoid EMI from UPS switching harmonics contaminating the quantum lab environment.

General room requirements. Clean, temperature-stable environment (20–24°C ± 2°C). Relative humidity 30–60% non-condensing. Positive air pressure relative to adjacent spaces (prevents dust and particle ingress during QPU swap operations). Restricted access via keycard or equivalent.

A full ISO cleanroom is not required. The QPU operates inside the dilution refrigerator’s vacuum cans, which are sealed against particulate contamination. What matters is a clean laboratory environment with controlled climate. Particle control is primarily important when the cryostat is opened for QPU installation or wiring tree maintenance.

Planning for the next system, not the current one

The single most common facility mistake in quantum computing deployments: building for the system you are installing today rather than the system you will need in three years. The incremental cost of over-specifying the facility at construction time is small compared to the cost of a retrofit when you upgrade from 20 to 100+ qubits or add a second cryostat.

Power headroom. A single 20-qubit system draws 10–15 kW steady-state. Specify at least 50 kW of 3-phase capacity, which accommodates a second cryostat, additional control racks, and a GPU node for NVQLink-based QEC decoding. Adding a 63 A panel during construction costs a fraction of what it costs to run new cable trays and panel boards after the facility is operational.

Cooling water headroom. Size the chilled-water plant for at least two pulse-tube compressors, even if you are installing one. Adding a second chiller unit after construction requires plumbing work in the quantum lab, which means a warm-up of the existing cryostat, which means downtime.

Floor space. The reference floor plan shows a single cryostat with surrounding clearance. For a two-cryostat facility, double the shielded lab footprint at design time. Adding a second shielded room after the first is operational is a major construction project. Two cryostats also enable redundancy: the IQM/LRZ facility operates two hybrid systems, so when one machine undergoes maintenance, the other continues serving users.

Helium recovery capacity. Size the gas-handling and recovery system for the total helium-3 charge across all planned cryostats, not just the first. Storage enclosure space is inexpensive; helium-3 cylinder storage is not something you want to improvise later.

Network infrastructure. Run more fiber than you need today. NVQLink requires 400 Gb/s Ethernet between the control electronics and the GPU node. Future quantum-to-quantum interconnects (photonic links between separate QPUs, trapped-ion networking between trap chips) will require dedicated optical fiber paths that are far easier to install during construction than as a retrofit.

Crane capacity. If you are installing an XLD-class system but might upgrade to KIDE-class within five years, rate the overhead crane for the heavier system. KIDE vacuum shields and radiation shields weigh substantially more than XLD components.

The general principle: the facility should be the most permanent element of the quantum computing deployment. QPUs are replaced on 2–5 year cycles. Control electronics are upgraded. Software changes continuously. The building and its infrastructure should support at least a decade of quantum hardware evolution without structural modification.

How facility requirements differ by modality

The parameters above are driven by superconducting transmon qubits, which have the most demanding facility requirements of any modality. Other approaches present a strikingly different picture.

Trapped-ion systems eliminate the dilution refrigerator and its associated floor loading, chilled-water, and helium-3 requirements. They replace it with a different set of demands: ultra-high vacuum chambers (smaller and lighter than a cryostat, but requiring days-to-weeks bake-out at ~200°C), optical tables (typically 1.5 × 3 meters, vibration-isolated, with Class 1000 cleanliness on the table surface), and extensive laser infrastructure (5–10 lasers per ion species, each requiring thermal stability to ±0.1°C for frequency stabilization). The vibration sensitivity is comparable to superconducting systems. The EMI requirements are less stringent (ions are not addressed by microwaves in most implementations, though microwave-based trapped-ion systems like eleQtron’s EGALE would have EMI requirements closer to superconducting). Safety considerations shift from helium asphyxiation hazards to laser safety: Class 4 lasers require interlocked rooms, beam shutters, emergency stop buttons, and laser safety training for all personnel with access. The floor layout is dominated by the optical table, laser racks, and beam delivery system rather than by a cryostat. Total facility footprint is comparable to superconducting but the cost of facility preparation is generally lower because no cryogenic cooling infrastructure or helium storage is required.

Neutral-atom systems have the simplest facility requirements of any modality. Pasqal’s Orion series fits in a standard 19-inch server rack, draws approximately 3 kW, and operates at room temperature. No dilution refrigerator. No helium-3 or helium-4. No chilled-water plant beyond standard rack cooling. No special floor loading requirements (a rack-mounted system distributes weight through standard rack feet). The vibration sensitivity exists but is managed by the system’s internal optical-table isolation. The primary facility concerns are laser safety (Class 3B or Class 4 lasers for trapping and Rydberg excitation, contained within the system enclosure but still requiring laser safety protocols for service access) and ultra-high vacuum management (contained within the system). A neutral-atom quantum computer can be deployed in a standard data center with standard single-phase power (15–20 A), standard air cooling, and standard floor loading. This is the modality that most closely resembles a conventional IT equipment installation. For organizations whose facility cannot be modified to accommodate superconducting-class requirements, neutral atom is the modality that can deploy with minimal or no facility remediation.

Silicon-spin systems require cryogenic infrastructure, but the higher operating temperature (approximately 1 K versus 10–20 mK for transmon) changes the helium equation. Bluefors’ XLDHe High Power system operates at 1–1.2 K using helium-4 only, eliminating the helium-3 dependency entirely. Helium-4 costs roughly $5–10 per liter, not $2,500. This removes the most expensive and strategically constrained consumable from the facility design. Floor loading, vibration, and EMI requirements are comparable to superconducting (the dilution refrigerator is similar in size and weight). The chilled-water requirement remains. The facility design is essentially the same as for superconducting, minus the helium-3 storage and recovery infrastructure. The Diraq + imec industrial fabrication path may eventually enable silicon-spin QPUs to be manufactured in standard CMOS fabs, but the cryogenic operating environment remains at the deployment site.

Photonic systems operate the core processor at room temperature on an optical table or in rack-mounted photonic modules. The cryogenic requirement is limited to the detector subsystem: superconducting nanowire single-photon detectors (SNSPDs) at 0.8–4 K or, for transition-edge sensors (TES), a small dilution refrigerator reaching 50 mK. PsiQuantum’s infrastructure uses cuboid racks fed by industrial cryoplants operating at 2–4 K, a qualitatively different setup from the dilution-fridge model. Floor loading is modest. Vibration isolation applies to the optical platform but is less stringent than for superconducting qubits. The dominant facility concerns are fiber management (photonic systems involve extensive optical fiber routing between source modules, processor modules, and detector modules), optical alignment stability (thermal stability requirements are tight), and the cryogenic detector subsystem’s cooling water and power. The total facility footprint can be comparable to superconducting for large-scale photonic installations (PsiQuantum’s Brisbane and Chicago compute centers are purpose-built facilities), but smaller installations resemble high-end optics labs more than cryogenic facilities.

The OCP quantum infrastructure workstream

The Open Compute Project launched its Future Technologies Initiative Quantum Information Infrastructure (QII) workstream at the OCP Global Summit in late 2025. The initiative brings together quantum hardware builders, data center operators, and HPC centers to define open specifications for co-locating quantum systems alongside classical infrastructure.

The first deliverable is the white paper “Integrating Quantum Processing Units into Data Center Infrastructure”, which establishes the community-led framework for defining physical and logical requirements for housing quantum systems in production data centers. A QPU Deployment Checklist accompanies the white paper. The IQM/LRZ deployment is the first case study shared through the OCP community.

OCP’s Chief Innovation Officer Cliff Grossner has indicated that OCP hopes to develop a quantum-ready certification for data center facilities, with a first draft expected sometime in 2026. This would extend OCP’s existing OCP-Ready self-assessment program (which evaluates data centers against best practices defined by hyperscalers like Meta and Microsoft) to include quantum-specific facility parameters.

For integrators and facility operators, the OCP workstream provides two things: a community-validated set of facility requirements that can be cited in procurement specifications and construction contracts, and a forward-looking standard that, once the quantum-ready certification materializes, will help data center operators market their facilities to quantum hardware buyers.

PostQuantum.com covered the OCP quantum initiative at launch: OCP Initiative to Integrate Quantum Computers into Data Centers.

The site survey checklist

A complete site survey for a quantum computer installation should cover every parameter listed in this article. The following checklist, adapted from Applied Quantum’s playbook and the OCP QPU Deployment Checklist, provides the specification-level detail:

Floor loading capacity (kg/m²) at the proposed cryostat location, verified by structural engineer sign-off. Vibration survey: minimum 25 hours of accelerometer data, evaluated against VC-A criterion or stricter, covering at least one full diurnal cycle plus one weekday rush-hour period. EMI spectrum analysis: 4–8 GHz band sweep, DC magnetic field mapping (target below 100 µT), AC magnetic field mapping (target below 1 µT), identification of all EMI sources within the 100-meter exclusion perimeter. Power quality: THD measurement (target below 1%), voltage sag and swell characterization, 3-phase balance verification, available amperage at the proposed panel location. Cooling infrastructure: chilled-water temperature (15–25°C required), available flow rate (10–30 L/min per pulse-tube compressor), available pressure (4–6 bar), routing path from chiller to cryostat. Ceiling height (minimum 3 meters clear). Door, corridor, and elevator clearances along the delivery path for cryostat crates (approximately 750 kg, XLD-class dimensions). Proximity to vibration sources: elevator shafts, rail lines, HVAC plants, loading docks, heavy machinery (100-meter exclusion zone). Proximity to EMI sources: cellular base stations, high-power broadcast antennas, high-voltage switchgear (100-meter exclusion zone). Helium storage: available space for secured, ventilated enclosure with oxygen sensors. Network connectivity: 10/25/100 GbE fiber availability to HPC infrastructure. Emergency power: UPS sizing for graceful shutdown (20-minute minimum at full load). Compressed air availability for the gas-handling system. Ambient temperature stability (20–24°C ± 2°C). Relative humidity (30–60% non-condensing).

The survey deliverable is a facility specification document that maps every parameter to a pass/fail assessment and provides remediation plans for any gaps. For most existing data centers, expect 2–6 remediation items. The most common are floor reinforcement, lighting replacement, and chilled-water plant installation. For purpose-built quantum labs, the survey validates the construction specifications before hardware procurement begins.

The facility as competitive advantage

Most coverage of quantum computing focuses on qubits, algorithms, and roadmaps. Facility preparation rarely makes the headlines. But in practice, the facility is often the longest-lead and most underestimated element of a quantum computer deployment. The IQM/LRZ team’s first lesson, documented in their arXiv paper, puts it directly: quantum computers have stricter facility requirements than classical systems, yet their deployment in HPC environments is feasible when preceded by a rigorous site survey.

The organizations that will deploy quantum computers fastest are the ones that start the facility work first, in parallel with hardware procurement, rather than waiting for hardware delivery to reveal that the floor cannot hold the weight, the lights create interference, or the cooling water is 20 degrees too warm.

For the specific cryostat, wiring, and QPU integration that happens inside the prepared facility, see the superconducting, trapped-ion, neutral-atom, and silicon-spin build guides. For the cryogenic infrastructure that dominates cost and timeline, see the dedicated cryogenics and helium-3 article. For the economics of the full deployment, see what a quantum computer actually costs.