Quantum computing is not "competing" with AI and HPC, but rather integrating with them... The key to popularization lies in software.

Some analyses point out that quantum computing, artificial intelligence, and high-performance computing (HPC) are not a “zero-sum” competition in which they are fighting for dominance. Instead, they are becoming “complements” that can deliver maximum effectiveness when used together. However, it is widely believed in the industry that the development environment is still in its early stages, and promoting quantum technology and improving software accessibility are regarded as core issues.

In a conversation with theCUBE at the HPE World Quantum Day event, Dave Bellante diagnosed that combining CPUs, GPUs, and quantum processing units (QPUs) to solve problems that have previously been difficult to tackle will become the core direction for next-generation technological innovation. Paul Gillin also emphasized that quantum computing software development today is essentially at a “primitive stage,” and that it needs a standardized development environment that anyone can easily master—something like a “quantum version of Python.”

Playing a “accelerator” role that supports existing supercomputers in the short term

Experts believe that the short-term impact of quantum computing is not about replacing existing workflows. Instead, it first manifests as an “accelerator” role—being able to execute certain parts of the computations handled by supercomputers faster. Tom Beck of the Oak Ridge National Laboratory explained that connecting quantum computers with HPC means that some calculations are handled by existing systems, while the most complex quantum-domain parts are handled by quantum devices. This is a realistic path.

The key lies in the speed and efficiency of information transfer between the two systems. This means quantum computing will not immediately change all computing environments; more likely, it will enter enterprise applications in a precise way—taking responsibility for specific problems within hybrid computing architectures.

The Argonne National Laboratory is also working to integrate quantum computing into practical research workflows in chemistry and materials science. Laura Schultz explained that in traditional HPC environments, quantum mechanical phenomena have to be achieved through simulation, whereas quantum computing can handle these problems more directly. The structure is: quantum devices handle calculations for specific intervals, then send the results back to simulation systems based on supercomputers to complete the remaining work.

Barriers to adoption are not hardware, but “engineering” and the software stack

Quantum computing has the potential to surpass existing supercomputers in processing large-scale and complex problems such as tracking neutrino behavior during interactions. Its commercial application potential in areas such as logistics optimization or new drug R&D has also been discussed for a long time. However, due to physical constraints and engineering challenges, the pace of real-world adoption has been slower than expected.

Kristy Beck of the Lawrence Livermore National Laboratory pointed out that in chemical problems that form the basis for drug interactions, quantum technology is expected to have significant effects. But the problems themselves are too complex, so commercial outcomes may appear later than in the logistics domain.

Amir Shehata of the Oak Ridge National Laboratory explained that in order to improve the accessibility of quantum technology, the entire technology stack must be redesigned. In particular, qubits’ operating conditions vary greatly depending on the hardware approach. The superconducting approach has a short lifetime and rapid degradation, requiring precise timing control; neutral-atom approaches have other limitations. This means that quantum computing software ultimately must be able to accommodate all these different hardware requirements.

He added that the new quantum software infrastructure may not consist solely of completely unfamiliar technologies, but will instead adopt forms that use familiar computing resources in the way people already use GPUs. This indicates that the adoption of quantum computing may follow a path connected to the existing AI-HPC ecosystem rather than one that is separate from it.

The key is “when to delegate which tasks to quantum processing”

There are also assessments saying that the true value of quantum computing is not in handling all problems, but in delegating the most suitable computational tasks at the best time. Thanks to superposition and entanglement, qubits can show advantages in complex mathematical problems that require reviewing multiple solution approaches simultaneously.

Mikael Johansson of CSC—the Finnish IT Center for Science—used the example of “green transformation” to point out that quantum computing could play an important role in developing better catalysts, next-generation batteries, and magnets. This means there is huge application space for quantum technology in industrial topics such as the energy transition and the development of advanced materials.

However, Dieter Kranzlmüller of Germany’s Leibniz Supercomputing Centre drew a boundary and said that quantum computers will not replace supercomputers. He explained that a more realistic approach is to establish an integrated structure in which systems automatically categorize tasks—sending some computations to supercomputers and others to quantum computers.

The Pawsey Supercomputing Research Centre in Perth, Australia, is also running the “Setonix-Q” project so that researchers can conduct quantum-mechanics experiments. Pascal Elahi said that the goal is not only to serve quantum researchers, but also to expand access for more users who want to solve real-world problems.

Although quantum computing has not yet reached the stage of full-scale popularization, it is rapidly expanding its potential for industries along a direction that combines with AI and HPC rather than replacing them. Ultimately, the market turning point may not lie in more powerful hardware itself, but in whether it is possible to build, as quickly as possible, a software environment and integrated infrastructure that more developers and researchers can use easily.

TP AI Notice: This article has been summarized using language models based on TokenPost.ai. The main content of the body may be omitted or may not be consistent with the facts.