Quantum Rings Data Shows How Users Run Quantum Computers

Discussions about quantum computing demand often rely on forecasts more than on concrete measurements. A new report from Quantum Rings aims to bridge this gap by analyzing the usage patterns on its Open Quantum platform during the summer of 2026. The resulting picture is specific and based on self-selected data, yet it represents one of the few cross-vendor records of actual quantum workloads publicly released.

Key Findings from the Open Quantum Report

Indeed, Quantum Rings operates Open Quantum, a routing layer that directs jobs to six quantum processing units from four hardware providers. The report, titled “Open Quantum Insights Edition 001” and released on September 1, 2026, analyzed Public Plan jobs completed between June 1 and August 31. Users on this plan agree to share execution data in exchange for free or discounted access.

This trade-off significantly influences the data. Private and confidential commercial jobs are excluded, which means the dataset primarily reflects educational, benchmarking, and early-stage research activities. Quantum Rings reported volume using only percentages and growth rates, without disclosing raw job totals. This makes it challenging to assess the statistical significance of any single number or to compare Open Quantum with other cloud services.

The study included hundreds of active users across 47 countries. Completed jobs increased by 115% from the spring period, whereas active users grew by 206%. The company explicitly stated that these numbers track the adoption of its own network, not the overall industry.

Circuit Widths Are Expanding at the High End

The median circuit still utilized six qubits over the summer, consistent with previous periods. However, significant movement occurred at the upper end of the spectrum. The 95th-percentile circuit expanded from 20 qubits in the spring to 96 qubits between June and August.

The proportion of jobs employing at least 50 qubits rose from 2% to 11%. Similarly, jobs using more than eight qubits increased from 14% to 39%. This suggests that a small group of users is pursuing larger-scale computations, whereas the majority continue to run smaller test circuits.

That said, a full state-vector simulation’s memory requirements double with each added qubit, which renders direct classical simulation impractical beyond a certain size. Still, other classical methods can approximate or exploit the structure of large circuits, so a high qubit count doesn’t automatically prove a quantum machine outperformed a classical one. The report focused solely on what users submitted, without measuring whether these circuits produced useful or correct results.

Price Increasingly Influences Job Placement

The most pronounced trend in the data concerns cost. Rigetti’s 108-qubit Cepheus-1 system handled 57% of all completed jobs, which is more than any other machine. Quantum Rings listed its price at $0.000425 per shot, the lowest on the network. A “shot” refers to a single execution of a circuit; due to the probabilistic nature of quantum measurements, users often repeat the same circuit multiple times.

Retail prices ranged from Rigetti’s rate up to 8 cents per shot on IonQ’s Forte-1, a staggering difference of approximately 188 times. Demand clearly concentrated at the lower-cost end. Users ran a median of 2,000 shots per job on Cepheus-1, compared to 100 on Forte-1, whereas the network median was 1,024 shots.

Nevertheless, cost wasn’t the sole determinant. IonQ’s higher-priced trapped-ion machines attracted a greater share of variational circuits, a class used in certain quantum machine learning and optimization tasks. Variational jobs constituted 29% of trapped-ion traffic, versus 11% across the entire network. Quantum Rings noted that trapped-ion systems accounted for a small portion of overall volume, which makes this split preliminary. 

Most Circuits Still Serve Hardware Testing Purposes

The report indicates that quantum computers are primarily being used as experimental instruments. Quantum Rings utilized a proprietary classifier to match circuit structures against known categories, such as Grover search, the quantum Fourier transform, QAOA, and machine-learning feature maps.

Among the circuits recognized by the classifier, 66% were designed for quantum state preparation or hardware benchmarking. Applications accounted for 20% and generally involved smaller circuits. Across all jobs, 37% didn’t match any known category, and this unclassified rate increased with circuit size, reaching 62% for circuits of at least 33 qubits.

The company referred to these larger, unmatched circuits as “bespoke,” and acknowledged that the classifier cannot distinguish between genuinely novel algorithms, modified benchmarks, or patterns its model failed to identify. There were also discrepancies between user-declared labels and the structural suggestions. For instance, only 22% of jobs labeled as “machine learning” were classified as quantum machine learning, and 34% appeared bespoke.

Short Waiting Times Challenge Common Beliefs

This dataset doesn’t support the notion that using a real quantum computer involves hours or days in a queue. Median time from submission to execution start was two minutes or less on all six systems. IonQ Forte Enterprise registered 38 seconds, whereas IQM’s Garnet and Forte-1 both had 60 seconds.

Quantum Rings posited that long waits arise from demand concentrating on a few popular machines, and are not an inherent characteristic of quantum hardware. A multi-vendor network can effectively distribute jobs. That said, there’s an obstacle: the measurement excluded time spent waiting for machines with limited schedules to become available, so it doesn’t encompass every user’s complete experience.

Conclusion

Bob Wold, founder and CEO of Quantum Rings, clearly articulated the report’s purpose. As he stated, emphasizing that Open Quantum serves as a cross-vendor record of workloads contributed by users who granted permission for their data to be studied and published:

“Quantum computing has no shortage of forecasts and no surplus of measurements.”

This study should be interpreted within its limitations: it’s a snapshot from one vendor, covering self-selected, primarily non-commercial traffic, without raw totals, and concluded on August 31, 2026. Within these bounds, it provides rare empirical evidence suggesting that price sensitivity is real, circuit sizes are gradually increasing at the high end, and queue times are not the commonly perceived barrier. The full report is available on the Open Quantum site. It’s advisable to hold these numbers loosely and observe whether the next edition confirms or reverses these trends.