A new study shows how quantum computing could help AI create more useful drug candidates and explore possibilities beyond its training data
Burnaby, British Columbia, 25 August 2026 – Generative artificial intelligence is quickly becoming an important tool in drug discovery, helping researchers design new molecules and explore potential treatments. However, AI systems can struggle when asked to create molecules that are both chemically valid and suitable for development as medicines. A new peer-reviewed study involving D-Wave and Japanese pharmaceutical company Shionogi suggests that quantum computing could help address this challenge.
The study, published in Scientific Reports, examined whether quantum annealing could improve the performance of a generative AI model used to design new drug candidate molecules. Researchers compared two versions of the same AI model. One relied entirely on classical computing, while the other used a D-Wave quantum computer as part of its sampling process.
The results showed a notable difference. The quantum-assisted model produced chemically valid molecules 97 percent of the time, compared with 73 percent for the equivalent classical model. It also produced a higher share of molecules considered drug-like. According to the study, 66.79 percent of molecules generated by the quantum model met the selected drug likeness standard, compared with 43.15 percent for the best classical model.
The findings are important because discovering a new medicine is a long and expensive process. Researchers must examine enormous numbers of possible molecular structures before identifying candidates that are safe, effective, and practical to manufacture. The study notes that the number of chemically possible drug-sized molecules is estimated to be greater than 10⁶⁰, while the number that can currently be synthesized and tested is far smaller.
Generative AI can help researchers explore this huge space by creating new molecular structures rather than simply searching through existing databases. However, models are trained on a limited amount of available data. This can cause them to produce molecules that look promising but fail when examined more closely.
This is where quantum computing could offer a new approach. In the study, the D-Wave quantum annealing system was used as a sampling engine inside the AI model. Instead of relying entirely on classical methods to estimate possible solutions, the researchers used quantum computing to explore complex combinations and generate samples from the model’s target distribution.
The researchers believe this capability could extend beyond pharmaceutical research. Generative AI systems in many industries face a similar problem. They need to create useful solutions in areas that may not be directly represented in their training data. Quantum computing could potentially help these systems explore the spaces between known examples and discover new possibilities.
The research therefore points to a broader connection between quantum AI and generative AI. Rather than replacing classical artificial intelligence, quantum computing could work alongside it, handling specific tasks where searching through large and complex possibilities is particularly difficult.
For the pharmaceutical industry, the potential impact could be significant. Better AI-generated drug candidates could help researchers reduce the number of unsuitable compounds that move into laboratory testing, saving time and resources during the early stages of drug discovery.
The study does not mean quantum computing has solved the challenges of drug development. More research and testing will be required to determine how these methods perform across different drug discovery tasks. Still, the results provide measurable evidence that quantum computing can improve certain generative AI outputs and could become an important part of the future AI and pharmaceutical technology landscape.
