How much does a quantum computer cost?
Disclosed and estimated prices for on-premise quantum computers range from tens of thousands of dollars for two-qubit educational systems to more than $100 million for flagship installations at national HPC centers. The median tracked deal is about $10 million. Price per qubit varies by modality, with annealing systems cheapest per qubit and photonic and trapped-ion systems most expensive. As of Q3 2026.
The price range
Across 175 tracked deals with a disclosed or estimated value, the median is $10.1 million. A quarter of deals fall below $4.95 million and a quarter sit above $26 million. The top decile starts at $43.6 million, and the largest single tracked deal is $145.7 million. At the bottom of the range, two-qubit desktop systems built for teaching sell for tens of thousands of dollars.
Two numbers therefore answer the question, depending on who is asking. A university buying a teaching machine spends less than $100,000. A national supercomputing center installing a flagship system spends between $10 million and $100 million.
Price per qubit by modality
| Modality | Systems priced | Average price | Average qubits |
|---|---|---|---|
| Superconducting | 11 | $26.1M | 74 |
| Trapped ion | 8 | $19.2M | 74 |
5 more rows in the dataset
Open the Quantum Computing Sales Dataset →Price per qubit is a trap. Annealing systems look cheap per qubit because they carry thousands of them, but an annealing qubit does a different job from a gate-model qubit. Room-temperature NMR and NV diamond systems are inexpensive because they are educational instruments, not computational machines. Compare within a modality, never across.
Sale, lease, or cloud access
Buying outright is the most expensive option and the most common in the data. Leasing spreads the cost and shifts obsolescence risk to the vendor, which matters in a field where hardware generations turn over every two to three years. Cloud access removes capital cost entirely and is how most organizations should start.
Institutions still buy on-premise systems for three reasons: sovereignty over sensitive workloads, low-latency coupling to a classical supercomputer, and the ability to run experiments a cloud service will not expose.
What drives the price
Qubit count matters less than the supporting stack. A superconducting system carries a dilution refrigerator, control electronics, and shielding, and those costs are close to fixed regardless of processor size. Trapped-ion and neutral-atom systems trade cryogenics for laser and vacuum infrastructure. Integration with an existing HPC facility, on-site support, and software licenses are frequently bundled into the headline figure, which is one reason published prices are hard to compare.
Where to find list prices
The Quantum Computing Sales Dataset includes a priced catalog of 75 systems from 32 vendors, each with the price type, a best estimate, a range, a confidence rating, and the evidence behind it, alongside every tracked deal with its value and the basis for any estimate.
Frequently asked
Rarely in the first years. Cloud access spreads cost over usage, while a system purchase adds installation, cryogenics or lasers, facilities, and staff. Buyers choose on-premise for sovereignty, latency, or research control.
Usually the processor, control electronics, and cooling or vacuum stack, with software and support bundled. The dataset's bundled items field records what each deal included.
Qubit quality, connectivity, and error rates differ so much between modalities that a qubit is not a standard unit. Use it within a modality, not across them.
Get the Quantum Computing Sales Dataset
Enter your work email and an analyst replies within one business day with the sample, the invoice, or an answer.