What Is a Quantum Application?
A quantum application is a real-world problem, in science or industry, that is addressed by running a quantum computation as one step in a larger workflow. It is not the same as a quantum algorithm: the algorithm is the procedure, while the application is the task a user needs done, such as estimating the energy of a molecule.
An application usually combines a problem statement, one or more algorithms, classical pre- and post-processing, and hardware able to run the computation. The term is only meaningful when each piece can be specified and the claimed cost, accuracy, and validation strategy can be compared with the best known classical approach for the same task.
That comparison connects to quantum utility, the question of whether quantum computers can produce useful results at a meaningful scale. It is also why quantum computing use cases need precise definitions rather than broad promises.
The Main Families of Quantum Applications
Most credible quantum applications fall into a few families, and the strength of the evidence differs between them.
| Family | Example problem | Status of the evidence |
|---|---|---|
| Simulating quantum systems | Molecules, materials, spin models | Proposed by Feynman in 1982; a natural fit, though strong approximate classical methods compete |
| Number theory | Factoring with Shor's algorithm (1994) | Superpolynomial speedup over known classical algorithms; needs fault-tolerant hardware |
| Search and optimization | Unstructured search, combinatorial problems | Grover's algorithm gives a proven quadratic speedup for unstructured search; no established large speedup for general optimization |
The first row covers quantum simulation and quantum chemistry, where the quantum computer mimics a system that is itself quantum.
Quantum machine learning is sometimes listed as a fourth family. Aaronson (2015) pointed out in "Read the fine print" that loading classical data and reading out results can erase a theoretical speedup, so it is best treated as open research.
What Makes a Quantum Application Credible
Credibility comes from resource estimates: how many logical qubits, how many non-Clifford gates, how much runtime, and what logical error rate the full computation needs. These numbers decide whether a problem fits a given machine. They depend on the overhead of quantum error correction and on how a quantum compiler translates the application into hardware operations.
Two published results show why this matters:
- Gidney and Ekerå (2021) estimated that factoring a 2048-bit RSA integer would take about eight hours using roughly twenty million noisy physical qubits, under stated assumptions about error rates and the surface code.
- Babbush and colleagues (2021) argued that quadratic speedups are unlikely to be enough on early error-corrected machines, because error-correction overhead and large constant factors can make the problem sizes and runtimes needed for advantage impractical. They urged attention to problems with larger speedups.
Estimates also move. Reiher and colleagues (2017) estimated the cost of simulating the nitrogenase cofactor FeMoco, and later algorithmic work, such as Lee and colleagues (2021), reduced the estimated cost of such chemistry calculations.
Quantum Applications on Neutral-Atom Hardware
Neutral atoms fit this topic naturally, because QuEra builds neutral-atom computers and the hardware maps directly onto some applications. Aquila, QuEra's analog system, has been on Amazon Braket since 2022. It runs analog Hamiltonian simulation: the user sets how Rydberg atoms are arranged and driven, and the system evolves under that Hamiltonian. This suits spin-model physics and unit-disk graph problems such as maximum independent set, and related embedded variants, which Ebadi and colleagues (2022) studied on Rydberg atom arrays as a research demonstration.
Gate-based applications need more. Gemini, QuEra's gate-based system, was installed at AIST in Japan as an on-premises system in 2025, and Libra, a fault-tolerant system, is planned for Amazon Braket in 2028. Bluvstein and colleagues (2024) demonstrated logical-level operations on a neutral-atom processor, an early step toward the error-corrected machines that chemistry applications require.
On the applications side, QuEra announced that its scientists will work with three of the nine teams in the ARPA-E Quantum Computing for Computational Chemistry (QC3) program, developing neutral-atom-based modeling frameworks for superconductors, high-performance magnets and hydrogen production. This is a research program, not a demonstrated result.
How to Evaluate a Claimed Quantum Application
A short checklist separates real applications from slogans:
- Which exact problem, and at what input size?
- What is the strongest known classical baseline, including approximate methods?
- Does the cost include data loading, compilation and readout?
- Does it need error correction, and if so how many logical qubits and gates?
- Is the claim a measured demonstration or a resource estimate?
FAQ
What is the difference between a quantum application and a quantum algorithm?
A quantum algorithm is a procedure, such as Shor's algorithm. A quantum application is the end-user problem, such as breaking a cryptosystem or computing a molecular energy, which may combine several algorithms with classical processing.
Are there quantum applications running today?
Research-scale applications run today, such as analog simulations of spin models on neutral-atom and other platforms. Most high-value applications, such as large chemistry calculations, are expected to need error-corrected hardware that does not yet exist at the required scale.
Which industries are named most often as users of quantum computing?
Chemistry and materials, energy, finance, logistics and cryptography come up most often. The evidence is strongest for simulating quantum systems and for factoring, and weaker for optimization and machine learning.
Is optimization a proven quantum application?
Not in general. Grover-type methods give a quadratic speedup for unstructured search, but many optimization approaches are heuristics without proven speedups. Babbush and colleagues (2021) argue a quadratic speedup may not yield practical advantage on early error-corrected machines once error-correction overhead is included.
Key Takeaways
- A quantum application is an end-user problem addressed with a quantum computation inside a larger workflow, while an algorithm is only the procedure.
- Simulation of quantum systems and factoring have the strongest theoretical case; optimization and machine learning remain less certain.
- Credibility rests on resource estimates and comparison with the strongest classical method, not on broad claims.
- Neutral-atom analog systems natively implement Rydberg Hamiltonians for analog simulation and unit-disk graph encodings, and the path to error-corrected applications runs through logical qubits.
.webp)
