What a Quantum Computer Actually Costs to Build and Operate

Introduction

The most common question from CTOs and laboratory directors evaluating on-premises quantum hardware is not “which qubit modality?” or “how many qubits?” It is: “What does this actually cost?”

The answer is more complex than a single number because a quantum computer is not a single purchase. It is a capital investment in hardware (cryostat, QPU, control electronics, wiring, calibration software, HPC integration), a facility investment (vibration isolation, EMI shielding, power conditioning, chilled water, helium recovery), an operational expenditure in personnel and consumables (3-8 specialized FTEs, annual cryostat maintenance, helium-3 management), and an ongoing commitment to vendor support contracts, QPU upgrades, and performance monitoring. The five-year total cost of ownership for even a modest 5-qubit research system exceeds the initial hardware purchase by a factor of two or more.

This article provides the line-item cost breakdown at three system tiers, the five-year and ten-year TCO models, the personnel and training costs that buyers consistently underestimate, and the government funding programs across 15+ countries that offset 50-80% of capital expenditure for qualifying organizations. For the hardware being costed, see the modality-specific build guides for superconducting, trapped-ion, neutral-atom, photonic, and silicon-spin systems. For the facility and cryogenic infrastructure costs that dominate the superconducting budget, see the dedicated articles.

Capital expenditure tiers

Four distinct system tiers serve different use cases, with costs that span three orders of magnitude.

Educational and desktop ($5K-$50K system cost, $20K-$100K annual operations). Cloud access credits or NMR desktop systems (SpinQ Gemini Mini at ~$5,000 for 2 qubits). Suitable for teaching and prototyping. No on-premises quantum hardware in the meaningful sense. Most organizations should start here before committing to hardware procurement.

Entry research ($1M-$3M system cost, $200K-$400K annual operations). A 5-9 qubit system: QuantWare Soprano-D5 QPU or Rigetti Novera, Bluefors LD450sl cryostat, Qblox Cluster or QM OPX+ control electronics, conventional coaxial wiring, Q-CTRL or QuantrolOx calibration software. Suitable for hardware familiarization, small algorithm development, and teaching graduate students to operate real quantum hardware. Five-year TCO: approximately $5 million.

Mid-range research ($3M-$15M system cost, $400K-$800K annual operations). A 17-64 qubit system: QuantWare Contralto-D21 or Tenor-D64 QPU, Bluefors XLD1000sl or Maybell Big Fridge cryostat, Cri/oFlex wiring, expanded control electronics (2-3 Qblox Cluster chassis), NVQLink GPU node for HPC integration. Suitable for algorithmic benchmarking, QEC research, and commercial R&D. Five-year TCO: approximately $10 million. Ten-year TCO: approximately $17 million.

Industrial ($15M-$500M+ system cost, $2M-$30M annual operations). A 100-1,000+ qubit system at national laboratory or hyperscaler scale. KIDE-class cryostat, industrial QPU (IQM Radiance, future QuantWare Baritone/VIO-40K), multi-rack control infrastructure, dedicated GPU cluster for QEC decoding, dual-cryostat redundancy. Euro-Q-Exa’s total acquisition cost was €25 million for a 54-qubit system with a 150-qubit upgrade. Five-year TCO: $30M-$80M. Ten-year TCO: $50M-$150M.

Where the money goes: line-item cost ranges

These are list-price estimates, not vendor quotes. Negotiated prices vary 20-40%.

QPU components. QuantWare Soprano-D5: ~€60K. QuantWare Contralto-D21: ~€300K. Rigetti Novera 9-qubit QPU only: ~$900K. Rigetti Novera complete system with upgrade packages: ~$2.85M (based on September 2025 purchase order data). QuantWare Tenor-D64: contact vendor. IQM and Quantinuum integrated systems: pricing not publicly disclosed, but the Euro-Q-Exa €25M figure for a 54-qubit IQM system plus HPC integration provides a reference point.

Cryostats. Bluefors LD450sl: $700K-$1M. Bluefors XLD1000sl: $1.5M-$2.5M. Bluefors KIDE: $5M-$10M class. Maybell Big Fridge: $1.5M-$3M. Oxford Proteox LX: contact vendor.

Control electronics. Qblox Cluster (entry configuration): $200K-$500K. Quantum Machines OPX+: $200K-$500K. Zurich Instruments QCCS: $200K-$500K. Lead times: 8-16 weeks, dependent on FPGA allocation.

Cryogenic wiring. Delft Circuits Cri/oFlex assemblies: varies by channel count, typically $50K-$200K for a mid-range system. Conventional coaxial wiring for a 5-qubit system: ~$30K.

Helium-3. Initial charge for an XLD-class system (~40 liters at $2,500/L): ~$100K. Recovery system (Quantum Technology Corp HR3/HRHP): ~$80K. Spare charge: ~$100K. This single line item ($280K for a mid-range system) is a cost that most first-time buyers do not anticipate.

Calibration software. Q-CTRL Boulder Opal Scale-Up or QuantrolOx Quantum EDGE: software license, typically $50K-$80K initial plus $30K-$50K annual. This is one of the highest-ROI line items in the budget because it compresses the bring-up timeline from months to weeks.

HPC integration. NVQLink GPU node (NVIDIA GH200 or GB200 class): $150K-$300K. Network infrastructure: $20K-$50K. QRMI/Slurm configuration is a labor cost, not a hardware purchase.

Facility preparation. Vibration isolation pad, EMI shielding, power conditioning, UPS, chilled-water plant, helium recovery: $200K-$400K for a typical retrofit of an existing building. Purpose-built quantum labs can cost more. See the facility preparation guide for the full parameter set.

Integration services. Applied Quantum or equivalent independent systems integrator: $200K-$400K for a mid-range system build, plus $50K-$100K annual support contract. The integration fee covers site survey, vendor selection, procurement management, multi-vendor interface engineering, commissioning, calibration automation deployment, and HPC integration. For organizations without in-house quantum engineering expertise, this is the difference between a five-month deployment (Q-PAC) and a twelve-month struggle.

The cost that surprises: personnel

Hardware vendors quote hardware prices. Facility contractors quote construction costs. Nobody hands the buyer a line item for “the five people you need to hire to operate this machine.” Yet personnel is the single largest cost category in the five-year TCO for every system tier.

Entry research system (5 qubits): 3 FTEs at an average loaded cost of $150K per year per person. Five-year personnel cost: $2.25 million, which is 44% of the $5.1 million five-year TCO. The roles: one cryogenic/facility engineer, one quantum control specialist, one HPC/DevOps engineer.

Mid-range research system (20 qubits): 5 FTEs. Five-year personnel cost: $3.75 million, which is 37% of the $10.1 million five-year TCO. Additional roles: a second quantum control specialist and a software engineer for framework integration and application support.

Production service with external users: 5-8 FTEs plus vendor support contracts. The additional roles: site-reliability engineer rotation (on-call, incident response, monitoring), additional software engineers for API development and user support.

For neutral-atom systems (Pasqal Orion), the personnel requirement drops to 2-3 FTEs plus vendor support because the vendor handles the laser, vacuum, and atom-loading subsystems through the support contract. For trapped-ion systems, the laser specialist role is the hardest and most expensive to fill. For superconducting systems, the cryogenic engineer is essential and cannot be shared with other responsibilities during the first year of operation.

The talent market for quantum operations personnel is tight. AMO physics PhD graduates who can maintain laser systems, condensed matter PhD graduates who can tune quantum dot gate voltages, and cryogenic engineers with dilution refrigerator experience are all in high demand and short supply. Budget for competitive salaries, training periods (3-6 months for a new hire to become productive on quantum hardware), and vendor training programs (QuantWare Academy, Q-CTRL Black Opal, IBM Qiskit Global Summer School).

Five-year TCO: entry research system

This model illustrates a 5-qubit superconducting system with a Bluefors LD450sl cryostat and a QuantWare Soprano-D5 QPU. All figures are approximate.

Year 0 capital expenditure: cryostat $850K, QPU $70K, control electronics $300K, wiring $30K, calibration software $50K, HPC integration $150K, helium-3 charge + recovery $110K, facility preparation $200K, integration services $200K. Total year-0 capex: approximately $1.96 million.

Annual operating expenditure (years 1-5): personnel $450K, calibration software license $30K, HPC integration maintenance $20K, cryostat maintenance $30K, utilities $25K, integration support $50K, helium-3 top-up (year 3) $15K. Total annual opex: approximately $610K.

QPU refresh (year 3): $70K for a replacement Soprano-D5.

Five-year TCO: approximately $5.1 million. Ten-year TCO: approximately $8.8 million (including a control electronics module refresh in year 5 and a second QPU refresh in year 7).

The ratio of year-0 capex to five-year TCO is roughly 1:2.6. For every dollar spent buying the hardware, expect to spend $1.60 operating it over five years. This ratio improves slightly at larger scales (personnel costs grow sub-linearly with qubit count) but never drops below 1:2.

The neutral-atom cost advantage

The facility preparation and cryogenic infrastructure costs that dominate the superconducting budget disappear for neutral-atom systems. A Pasqal Orion deployment at 140+ qubits eliminates: the cryostat ($700K-$2.5M), the helium-3 charge and recovery system ($180K-$280K), the chilled-water plant ($50K-$100K), the vibration isolation slab ($50K-$100K), the EMI shielding ($50K-$150K), and the cryogenic engineer role ($150K/year). The facility preparation cost drops from $200K-$400K to near zero for a standard data center with adequate rack space, power, and cooling.

Pasqal does not publicly disclose system pricing. But the infrastructure cost differential between a superconducting installation (where cryogenics, facility, and helium account for $1.5M-$3.5M of the year-0 capex) and a neutral-atom installation (where these costs are minimal or absent) is one of the strongest arguments for neutral-atom systems in the procurement decision. The operational cost differential (no helium management, no annual pulse-tube service, no unplanned warm-up risk) compounds over the system’s lifetime.

Government funding: who pays for quantum computers

Government programs worldwide are the primary funding source for on-premises quantum hardware procurement, particularly at the mid-range and industrial tiers. Many programs fund 50-80% of capital expenditure through competitive grants, co-investment arrangements, or direct procurement.

United States. DOE National Quantum Initiative and QIS Research Centers (five centers, tens of millions per center). DOE announced $65 million for quantum computing research in April 2026. DARPA QBI (Quantum Benchmarking Initiative, stages A/B/C, with performers including Riverlane, PsiQuantum, Xanadu, Diraq, IonQ, Quantum Motion, Atlantic Quantum). DARPA US2QC (Underexplored Systems for Utility-Scale Quantum Computing). DOD Quantum Applications Program ($59.5M in FY2026 NDAA). NSF Quantum Leap Challenge Institutes (~$25M per institute, five institutes). Quantinuum received a CHIPS R&D letter of intent for federal funding in May 2026. CSIS recommended the DOE establish multi-year procurement authority and budget for quantum computing services.

European Union. EuroHPC JU has deployed six quantum systems across European supercomputing centers: Euro-Q-Exa at LRZ (IQM, €25M), EuroQCS-Italy at CINECA (Pasqal), plus sites at BSC, CSC Finland, PSNC Poland, and Poznan. The SUPREME consortium secured €25M EU funding (€50M total with national co-funding) for industrializing superconducting quantum technologies across 23 partners in eight member states.

United Kingdom. The National Quantum Technologies Programme commits £2.5 billion over ten years (2024-2034). An additional £2 billion was announced in March 2026 for quantum computing procurement and scaling, including the ProQure program for prototype evaluation and integration into national computing infrastructure.

Germany. €3 billion action plan for quantum technologies. Bavarian State Ministry (StMWK) co-funds Munich Quantum Valley and Euro-Q-Exa. Jülich Supercomputing Centre hosts DGX Quantum and Pasqal deployments.

France. €1.8 billion National Quantum Strategy (2021-2025, renewed). Pasqal, Quandela, Alice & Bob, GENCI deployments.

Netherlands. Quantum Delta NL €615 million total program. Funds the Delft ecosystem (QuantWare, Qblox, Delft Circuits, QuTech HectoQubit/2).

Australia. PsiQuantum Brisbane compute center backed by AU$940M from federal and Queensland state governments. Diraq received AUD $20M from the National Reconstruction Fund Corporation. Q-CTRL headquartered in Sydney. National Quantum Strategy and NQIC.

Other significant programs. India’s National Quantum Mission (₹6,003 crore, approximately $700M, 2023-2031). Singapore National Quantum Strategy (S$300M+, including Quantinuum Helios deployment). Canada (C$360M National Quantum Strategy, CAD $23M for Xanadu’s Quantum Champions Program, negotiations for up to CAD $390M for Xanadu’s Project OPTIMISM). Japan (MEXT/AIST G-QuAT center with 18 KIDE systems, NEDO supply chain investment with QuEra). South Korea (₩3 trillion, approximately $2.2B, announced 2024, SDT Kreo 20-qubit system).

Applied Quantum assists clients with grant applications and co-development proposals as part of procurement engagements.

When cloud access is enough

Not every organization needs on-premises quantum hardware. Cloud access makes economic sense when your workload is below 10⁶ shots per month sustained, when no classified or regulated data prohibits cloud transit, and when you do not have a strategic imperative (sovereignty, talent attraction, sector leadership) that requires physical hardware.

Cloud pricing in 2026: IBM Quantum free tier provides access to 127-qubit systems. AWS Braket charges per shot and per task ($0.30-$0.75 per task plus per-shot charges depending on the backend). Azure Quantum provides credits for IonQ, Quantinuum, and Pasqal backends. OVHcloud offers Pasqal and Quandela access through its Quantum Platform.

For most organizations entering quantum computing, the right first step is $20K-$100K of cloud access to characterize workloads and identify use cases before committing to the $3M-$15M capital investment of on-premises hardware. The cloud tier is also the right validation step before writing a grant application: demonstrating a use case on cloud hardware strengthens any procurement proposal.

Operational costs that surprise first-time buyers

Beyond personnel, several operational cost categories consistently catch first-time quantum computer operators off guard:

Helium-3 management. The initial charge is budgeted in capex. The recovery, purification, and top-up costs after each warm-up event are opex. Budget $15K-$30K per warm-up for helium-3 handling. With at least one planned warm-up per year and the possibility of unplanned events, annual He-3 opex runs $15K-$50K depending on system reliability.

Unplanned downtime cost. An unplanned warm-up (power outage, compressor failure) costs 5-10 days of recovery. At $10K-$50K per day in personnel idle time, lost user access, and schedule disruption, a single event is a five- to six-figure incident. The $80K-$150K spare parts inventory recommended in the cryogenics article is insurance against this cost.

QPU refresh cycle. Transmon QPUs are not lifetime components. As the field advances, QPUs that were state-of-the-art at installation become obsolete within 2-4 years. QuantWare’s upgrade path (Soprano to Contralto to Tenor, same form factor) enables QPU upgrades without cryostat replacement, but each upgrade costs €60K-€300K+ for the new chip plus 5-10 days of downtime for warm-up, swap, cool-down, and recalibration.

FPGA allocation delays. All three major Western control platforms (Qblox, Quantum Machines, Zurich Instruments) rely on AMD/Xilinx and Intel/Altera RFSoC FPGAs. These compete for allocation with defense, telecom, and AI workloads. If allocation extends past 16 weeks, the control electronics become the schedule bottleneck. The hedge: pre-qualify FPGA allocation before ordering, confirm delivery timeline in writing, and consider maintaining a relationship with two control vendors for flexibility.

Software license renewal. Calibration software (Q-CTRL, QuantrolOx) and framework licenses (vendor-specific tools) run $30K-$80K per year. These are recurring costs that are easy to overlook in a capex-focused procurement plan.

What this means for procurement

Three principles from the TCO data:

First, budget the full TCO from day one. The five-year cost of operating a quantum computer is 1.6-2.6 times the initial hardware purchase. A $4.4M mid-range system purchase becomes a $10.1M five-year commitment when personnel, maintenance, consumables, and support are included. If the full TCO exceeds the available budget, start with cloud access rather than an undersized on-premises system.

Second, apply for government funding before ordering hardware. The programs listed above can offset 50-80% of capital expenditure for qualifying organizations. A €25M Euro-Q-Exa system with EuroHPC co-funding may cost the host institution €10M or less out of pocket. The grant application timeline (3-12 months) should run in parallel with the facility preparation described in the facility guide, not after it.

Third, compare modalities on TCO rather than qubit count or gate fidelity alone. The superconducting modality has the most mature QOA supply chain and the best-characterized performance, but it also has the highest facility and operational costs. Neutral-atom systems eliminate the entire cryogenic infrastructure cost layer and require fewer operational personnel. For organizations where the facility cannot be modified to accommodate superconducting-class requirements, neutral atom may be the only viable on-premises option at any budget.

For the Quantum Open Architecture model that enables component-level procurement and the HPC integration that connects the quantum computer to classical compute resources, see the corresponding articles in this series.