Quantum Hardware Costs More to Operate Than to Buy

Ask a vendor what a quantum computer costs and you get a hardware price. Ask someone who runs one and you get a number two to three times larger. For a 5-qubit superconducting research system, Applied Quantum’s Systems Integration Playbook (v2.0) puts year-0 capital expenditure near $1.96 million and annual operating expenditure near $610,000. Five years in, the running total is about $5.1 million. Roughly $1.60 of operations follows every dollar of hardware.

That gap is where quantum procurement decisions go wrong. The capital request gets approved, the machine arrives, and the operating budget turns out to be the real commitment.

The ratio holds at every tier

Capital expenditure is what you pay to acquire and install the system. Operating expenditure is what you pay each year to keep it producing results. Across the tiers we see in procurement work, the relationship between them is stable enough to use as a planning rule.

An entry research system runs about 1:2.6, capex to five-year total. A mid-range 20-qubit installation is close behind: roughly $4.4 million of purchases becomes about $10.1 million over five years. The ratio improves slightly as systems grow, because headcount rises more slowly than qubit count, but it doesn’t drop below 1:2 at any tier we have modeled.

For a board paper, that gives you a defensible first pass. Take the quoted hardware price, multiply by 2.5, and present that as the five-year figure. If the multiplied number exceeds the budget, the honest conclusion is that the system is out of reach, not that the operating costs will be managed down later.

All the ranges below are list-price estimates from the same playbook. Negotiated prices move 20 to 40 percent, and every figure here should be treated as a planning input rather than a quote.

Payroll is the largest single line

No vendor quotes the people. Yet in every model we build, salaries are the biggest category.

An entry system needs three full-time equivalents: a cryogenic and facility engineer, a quantum control specialist, and a high-performance computing (HPC) or DevOps engineer. At a loaded cost near $150,000 each, that’s $2.25 million over five years, or 44 percent of the total. A mid-range system needs five, at $3.75 million and 37 percent. A production service with external users needs five to eight, once you add on-call rotation and API support.

Two things make this harder than an ordinary hiring plan. The candidate pool is small, particularly for cryogenic engineers with dilution-refrigerator experience and for laser specialists on trapped-ion systems. And a strong new hire still takes three to six months to become productive on live quantum hardware, which means the machine’s first months are funded, staffed, and only partly useful.

Cryogenics is a decision, not a given

Superconducting systems live inside a cryostat, the refrigeration stack that holds the processor near absolute zero. That stack sets much of the budget. The refrigerator itself runs $700,000 to $2.5 million depending on class. The initial helium-3 charge and its recovery system add $180,000 to $280,000. Then come the chilled-water plant, the vibration isolation slab, the electromagnetic shielding, and a cryogenic engineer who cannot be shared with other work during the first year.

Neutral-atom systems remove most of that layer. There’s no dilution refrigerator, no helium-3 inventory, no annual pulse-tube service, and no unplanned warm-up risk. Facility preparation drops from a retrofit project to a question of floor space, power, and cooling in an existing HPC room. Headcount typically falls to two or three, because the vendor handles the laser, vacuum, and atom-loading subsystems under support.

Pasqal and others in that modality don’t publish system pricing, so the comparison can’t be closed on purchase price alone. What can be compared is the infrastructure layer, and on that layer the difference is measured in millions. This is the argument for putting modality choice inside the financial model rather than treating it as a separate technical decision made upstream.

The costs that arrive after commissioning

Five categories catch first-time operators, and all five are recurring.

Helium-3 handling. The initial charge sits in capex. Every warm-up after that costs $15,000 to $30,000 in recovery, purification, and top-up. With one planned warm-up a year, annual spend runs $15,000 to $50,000.

Unplanned downtime. A power outage or compressor failure means five to 10 days of recovery. At $10,000 to $50,000 a day in idle staff, lost access, and schedule slip, one event is a six-figure incident. A spare-parts inventory is insurance against it.

Processor refresh. Quantum processing units are not lifetime components. A QPU that leads the field at installation is typically superseded within two to four years. Upgrade paths that keep the same form factor avoid a cryostat replacement, but each swap still costs a new chip plus five to 10 days of warm-up, exchange, cool-down, and recalibration.

Control electronics lead times. The major control platforms all depend on radio-frequency (RF) system-on-chip field-programmable gate arrays (FPGAs) that compete for allocation against defense, telecom, and AI demand. Quoted lead times run eight to 16 weeks. Pre-qualify allocation before ordering, and get the delivery date in writing.

Software renewals. Calibration and control software runs $30,000 to $80,000 a year. It’s also among the highest-return items in the budget, because automated calibration compresses bring-up from months to weeks.

Two ways to spend the same $2 million

Take a fixed $2 million and put it against an entry research system. It covers the year-0 capital expenditure and leaves nothing for year one. The team then asks for $610,000 a year in a budget cycle where the capital request has already been spent.

Put the same $2 million against cloud access instead. Sustained cloud usage for a characterization program runs $20,000 to $100,000 a year, and two engineers to design and interpret the work cost roughly $300,000. That’s about $400,000 a year, or five years of continuous access and staffed effort for the same money, with no facility retrofit and no helium inventory.

Cloud access is the better answer when three conditions hold: sustained demand below about a million shots a month, meaning individual circuit executions; no data-residency or classification rule that forbids sending workloads offsite; and no sovereignty, talent, or sector-leadership reason that requires the machine to be physically yours. Public providers currently offer free tiers on large systems and per-task pricing in the range of $0.30 to $0.75 plus per-shot charges, which makes a characterization program cheap relative to what it tells you.

There’s a sequencing benefit too. A workload characterized on cloud hardware gives the procurement request a measured cost and performance baseline instead of a projection.

Grants change the arithmetic, if the timing works

Public funding is the primary source for on-premises quantum hardware at the mid-range and industrial tiers, and programs commonly cover a substantial share of capital expenditure through competitive grants or co-investment. EuroHPC has co-funded quantum systems at supercomputing centres across Europe. The UK’s National Quantum Technologies Programme and India’s National Quantum Mission both run on multi-year national footings.

The trap is sequence rather than eligibility. Application-to-award cycles run three to 12 months. That clock has to run in parallel with site survey and facility preparation, not after the equipment has been ordered. Teams that order first and apply second end up funding the gap themselves.

What belongs in the business case

• The five-year total, not the purchase price, as the headline number.
• A headcount plan with named roles, loaded costs, and a three to six month ramp before productivity.
• A modality comparison run on total cost, with the cryogenic infrastructure layer priced separately.
• A recurring line for helium, refresh, and software renewal, rather than a single contingency percentage.
• A funding timeline that starts before the purchase order.

Where to build the skill

Costing a quantum system well is a procurement discipline before it is a physics one. It asks for the ability to read a vendor specification against an operating model, to price a facility layer, to see the personnel implication of a modality choice, and to write a case that survives five annual budget reviews.

That’s the ground our systems integration and procurement material covers, and it’s built for the people who have to sign the paper rather than the people who tune the hardware. Enroll in the Systems Integration and Procurement certification at Quantum Academy. For the component-level engineering behind these numbers, the original cost and procurement analysis on PostQuantum.com goes deeper into the hardware itself.