Quantum AI refers to research that brings together quantum computing and artificial intelligence. The central question is whether computers that use the rules of quantum mechanics can run learning, optimization or simulation methods faster or better than classical computers. The best-known branch is quantum machine learning. The field is young and mixes real theoretical results with considerable hype, and today’s hardware is too small and noisy to show clear practical advantages for most AI tasks. It relates to machine learning, neuromorphic computing and AI hardware.
Background: quantum computing in brief
A classical computer stores information in bits that are either 0 or 1. A quantum computer uses qubits, which can be in a superposition of states, and operations on several qubits can create entanglement, a correlation with no classical equivalent. A quantum algorithm arranges these effects so that wrong answers cancel and right answers become more likely when the qubits are measured. Quantum computers are not simply faster at everything. In 1994 Peter Shor described an algorithm that could factor large integers efficiently, showing that quantum machines could in principle beat classical ones on certain problems [1]. For most tasks, no such advantage is known.
How quantum and AI meet
### Quantum machine learning
Researchers study several approaches:
- Quantum data with quantum algorithms, for example learning properties of quantum systems.
- Classical data with quantum algorithms, where data are encoded into qubits and processed by a quantum circuit. Encoding data efficiently is a major obstacle.
- Hybrid methods, in which a quantum circuit with adjustable parameters is trained by a classical optimizer, called a variational quantum circuit or quantum neural network.
- Quantum kernels, which use a quantum computer to compute similarity between data points for use in methods such as support vector machines.
### Quantum optimization
Many AI problems involve optimization, such as training models or scheduling. Methods like the quantum approximate optimization algorithm and quantum annealing target such problems, though evidence of practical speedups is limited and debated.
### AI for quantum computing
The relationship also runs the other way. Machine learning helps calibrate quantum devices, decode errors, design circuits and discover new quantum experiments. Some of the most convincing near-term links between the fields are of this kind.
### Quantum simulation for science
Quantum computers may one day simulate molecules and materials accurately, which could generate high-quality data for AI models in chemistry and materials science; see AI in scientific research.
The state of the hardware
Current devices are described as NISQ, for noisy intermediate-scale quantum, a term introduced by John Preskill for machines with tens to a few hundred noisy qubits that lack full error correction [3]. In 2019 Google researchers reported that a 53-qubit processor completed a sampling task they estimated would take a classical supercomputer thousands of years, though other researchers argued that improved classical methods could do it far faster [2]. Such experiments show progress but use tasks designed to favor quantum devices, not useful AI workloads. Large-scale, fault-tolerant machines with many thousands of reliable qubits are expected to need years of further development.
Possible benefits
- Speedups for particular learning or sampling problems, if they can be shown.
- Better models of quantum systems in chemistry and physics.
- Alternative ways to represent complex probability distributions.
- Improved optimization for certain structured problems.
Challenges and criticisms
- Data loading. Putting large classical data sets into a quantum state can cancel out theoretical speedups.
- Noise and error correction. Errors accumulate quickly in current hardware.
- Trainability. Some variational circuits suffer from barren plateaus, regions where gradients vanish and training stalls.
- Classical competition. Some claimed quantum speedups in machine learning were later matched by new classical algorithms, a process sometimes called dequantization.
- Cost and access. Hardware is expensive and mainly available through cloud services.
- Hype. Claims that quantum computers will soon transform AI are not supported by current evidence; see AI hype.
- Security. Large quantum computers could break widely used encryption, which has driven work on post-quantum cryptography; see AI cybersecurity.
Outlook
Most experts expect quantum AI to be useful first in specialized scientific tasks and as a tool for improving quantum hardware itself, rather than as a general replacement for graphics processors in training large models. Progress will depend on better error correction, larger machines and clear demonstrations on problems that matter. Meanwhile, companies, universities and governments invest in software frameworks, education and standards so that researchers can test ideas on simulators and small devices; see AI governance.
A worked example
A chemistry team wants to predict how a candidate catalyst binds a small molecule. Classical methods struggle when electrons interact strongly. The team writes a variational quantum circuit that prepares a trial state of the molecule and uses a classical optimizer to adjust its parameters so the energy is as low as possible. On today’s small, noisy hardware, they can handle only a tiny model and must correct for errors, so the result matches classical calculations rather than beating them. They record the circuit depth, error rates and runtime, then compare with classical baselines. The experiment shows how the method works and what hardware improvements would be needed for real advantage.
Software tools and learning paths
Several open-source frameworks let people write quantum circuits and run them on simulators or on cloud-hosted devices, and some integrate with standard machine learning libraries so that a quantum circuit can be treated as a layer in a model. Learners usually begin with linear algebra, probability and basic quantum mechanics, then move to circuits, variational algorithms and simple classification tasks. Because simulating more than a few dozen qubits on a classical computer is costly, most educational examples use small problems, which is useful for understanding methods but not for judging practical advantage.
Further reading and next steps
Readers can continue with neuromorphic computing, probabilistic AI and AI in scientific research to compare other paths toward faster or more efficient learning. When reading news about quantum breakthroughs, it helps to ask what task was solved, how it compares with the best classical method, how many qubits and what error rates were used and whether the result has been independently checked.
FAQs
What is quantum AI? It is research that combines quantum computing with artificial intelligence, mainly quantum machine learning and AI methods for improving quantum computers.
Do quantum computers make AI faster today? Not in practice. Current devices are small and noisy and have not shown clear advantages for real AI workloads.
What is a qubit? It is the basic unit of quantum information, which can exist in a superposition of 0 and 1 states.
Will quantum computers train large language models? That is not expected in the near term, because data loading and hardware limits are major obstacles.
Can I try quantum machine learning? Yes. Cloud services and open-source simulators let learners run small circuits.
More in Emerging AI Research
Key facts
- Peter Shor published a quantum algorithm in 1994 that could factor large integers efficiently, which showed that quantum computers could in principle outperform classical ones on certain problems1.
- In 2019 Google researchers reported that a 53-qubit superconducting processor completed a sampling task that they estimated would take a classical supercomputer thousands of years, a claim that other researchers disputed2.
- John Preskill coined the term NISQ, for noisy intermediate-scale quantum, to describe today's devices with tens to a few hundred noisy qubits that lack full error correction3.
Related concepts
Quick checkWhat is quantum AI?Show answer
Quantum AI is the study and development of methods that combine quantum computing with artificial intelligence, such as quantum machine learning algorithms, to explore whether quantum hardware can speed up or improve learning, optimization and simulation tasks.
Sources
- Shor, Algorithms for quantum computation: discrete logarithms and factoring, Proceedings of the 35th Annual Symposium on Foundations of Computer Science, https://ieeexplore.ieee.org/document/365700
- Arute et al., Quantum supremacy using a programmable superconducting processor, Nature, https://www.nature.com/articles/s41586-019-1666-5
- Preskill, Quantum Computing in the NISQ era and beyond, Quantum, https://quantum-journal.org/papers/q-2018-08-06-79/
Editorially reviewed by Specialty Digest Editorial TeamLast reviewed November 17, 2026Researched and drafted with AI assistanceReport an issue
