Quantum and AI begin to converge in hybrid computing experiments

Share Copy Link Share on X Share on Linkedin Share on Facebook Quantum computing and AI converge to accelerate development of both technologies. Photo credit: Gorodenkoff via Shutterstock As artificial intelligence (AI) systems grow larger and more expensive to run, the technology industry is searching for new ways to improve efficiency. That pressure is drawing renewed attention to quantum computing. Rather than focusing only on the long-standing idea that quantum computers might one day outperform classical machines, researchers are beginning to explore how the two technologies could complement each other. AI tools are already helping developers design and optimise quantum software, while some researchers believe quantum processors could eventually act as accelerators for certain AI workloads.One driver for renewed interest in quantum ideas is the growing pressure on AI infrastructure. Large language models and other generative AI systems require enormous computational resources. Training and running these models can require vast amounts of memory, energy and specialised hardware, pushing organisations to explore ways to make AI systems more efficient. Some companies are already applying techniques developed in quantum research to tackle that challenge. Multiverse Computing, for example, has developed an AI model compression technology based on tensor networks, mathematical tools created to simulate quantum systems. The approach analyses the internal structure of neural networks to identify which parameters are most important for specific tasks. “Especially in large language models, you have massive matrices with a lot of parameters,” says David Montero, director of AI research at Multiverse Computing. “By analysing correlations between inputs and outputs, we can understand which weights are important and which are not,” he adds. The company says this allows redundant parameters to be removed while preserving most of a model’s capabilities, an approach it argues is becoming increasingly important as AI moves beyond large data centres and into edge devices. “We’re talking with customers across industries who all have the same problem,” Montero says. “AI models today are not designed to fit into devices. Energy consumption and memory requirements become major constraints.” AI helping developers build quantum software While quantum-inspired mathematics is already influencing AI engineering, the reverse relationship is also beginning to take shape. A growing number of companies are exploring how AI tools might help accelerate the development of quantum software itself. There is no clear line of sight yet to commercial-scale quantum computing. However, it is likely to begin by 2027, according to soundings taken by research and analysis company GlobalData among quantum industry executives and corporates readying themselves to deploy the technology when feasible. One of the biggest barriers to practical quantum computing today is the complexity of programming quantum systems, which requires highly specialised expertise.Classiq, a quantum software company, is using AI to simplify that development process. It recently introduced an AI-assisted coding interface designed to help developers generate and refine quantum programs more quickly. At Nvidia’s GTC conference, it presented a broader strategy built around hybrid computing stacks that combine classical AI systems with quantum processors.“The main thing we’re showing is what AI can do for quantum coding,” says Simon Fried, vice president of corporate communications at Classiq. “It’s essentially about the front end of development—helping developers get up and running much faster.” The system allows developers to describe a problem in natural language or reference an academic paper, after which the platform generates a quantum program based on that description. “It’s not that different from the co-pilot tools people are used to in classical development,” Fried says. “You’re basically not coding anymore. You’re editing.” The approach could significantly reduce the time required to prototype quantum applications. Even experienced developers within the company now start projects using the AI interface, Fried says, because it allows them to move from concept to working code far more quickly. “What might previously have taken days or weeks of experimentation can now be done in minutes or hours,” he adds. Bridging the quantum skills gap AI-assisted development could also help address the persistent skills gap in quantum computing. Quantum programming remains a highly specialised discipline, and organisations often struggle to find developers with the required expertise. AI tools could make it easier for domain specialists—such as chemists, physicists or financial analysts—to contribute to quantum projects without needing deep knowledge of quantum programming. “If you have a domain expert who can clearly express the problem they want to solve, the AI can help translate that into quantum code,” Fried says. “That allows them to engage with quantum teams where previously they might have been speaking completely different languages.” The technology is not a complete replacement for experienced quantum engineers. AI-generated code still needs to be reviewed and refined by experts, particularly for complex applications. “It’s very similar to what we see in classical AI coding,” Fried says. “Your first result may be usable a large percentage of the time, but there will always be cases where you need experts to check and improve the output.” Hybrid computing architectures emerge The intersection of AI and quantum computing is also influencing how organisations think about computing infrastructure. Rather than replacing classical systems, quantum processors are increasingly being positioned as specialised components within larger hybrid computing environments. Many quantum algorithms are first developed using classical simulations before being deployed on quantum hardware, with GPUs well-suited to the complex linear algebra involved. Companies such as Nvidia are investing heavily in software environments designed to support these hybrid workflows. Classiq has integrated its development platform with Nvidia’s CUDA-Q environment, enabling quantum algorithms designed on its platform to run inside GPU-based simulation environments. The goal is to make the transition from development to execution as seamless as possible, says Fried. “It’s really about speed and flexibility,” he says. “How quickly can you go from generating an idea with AI to actually running the simulation and seeing the results?” The broader computing infrastructure that could support this kind of hybrid workflow is also beginning to take shape. IBM recently published a reference architecture for what it calls “quantum-centric supercomputing,” describing how quantum processors could be integrated into high-performance computing environments alongside GPUs and CPUs. Today, hybrid quantum-classical workflows are often assembled manually by researchers, who must coordinate job scheduling, data transfers and orchestration between systems. “Our architecture shows how quantum and classical resources can interact across application, orchestration and execution layers,” says Antonio Córcoles, head of quantum and HPC at IBM Quantum. “Today’s hybrid workflows are still fragmented and often assembled by hand.” The proposed architecture combines quantum processing units with classical compute clusters, high-speed networking and shared storage, enabling hybrid algorithms to divide tasks between quantum circuits and classical processing.Researchers are already experimenting with these systems in areas such as molecular simulation and materials science. Quantum computing as a potential AI accelerator While AI tools are already helping quantum developers today, the reverse relationship—quantum systems accelerating AI workloads—remains largely experimental. Quantum processors could eventually improve parts of machine learning pipelines, says Richard Murray, co-founder and CEO of photonic quantum computing company ORCA Computing. “We see quantum as an accelerator for AI,” Murray said at a recent industry event in London. The idea is not to replace existing AI infrastructure but to augment it. “You take what you’re already doing—GPUs, classical models—and you put a quantum system alongside them,” said Murray. Quantum systems have unique mathematical properties that may allow them to handle certain computational tasks more efficiently than classical machines. In particular, quantum processors are well-suited to problems involving correlations between large datasets. Those properties could potentially be useful in areas such as optimisation, sampling and generative modelling. However, researchers caution that the path to practical applications is far from straightforward. “Quantum plus AI sounds easy, but it’s really hard to do in reality,” Murray said.Early experiments point toward hybrid AI-quantum systems For now, the intersection between AI and quantum computing remains largely exploratory. AI tools are already helping developers work more effectively with quantum systems, while quantum ideas are influencing how AI models are designed and optimised. The possibility that quantum processors could eventually enhance AI workloads adds another layer of long-term potential. Rather than competing technologies, the two fields may evolve together as part of a broader shift toward hybrid computing architectures. As organisations continue to push the limits of AI infrastructure, that convergence could become one of the most active areas of experimentation in next-generation computing. Sign up for our regular news round-up! Give your business an edge with our leading Tech Monitor Sign up