Researchers at the Technical University of Denmark have shown that a quantum computer can improve the accuracy of a generative AI drug-discovery model, using a hybrid workflow that paired their protein-prediction system with a printer-sized quantum machine built by British startup ORCA Computing. The team used the setup to generate novel peptides — short chains of amino acids designed to bind to specific proteins in the body, a foundational step in vaccine development. When the peptides were synthesized and tested in the lab, the hybrid model produced more successful binders than its classical counterpart, with the largest improvements on targets where training data was scarce.
The work is notable less for the size of the compute involved than for the fact that it produced a lab-verified result at all. Quantum computing has struggled for years to demonstrate near-term commercial usefulness, and most claims of quantum advantage have lived inside benchmarks rather than wet-lab experiments. The DTU team ran a generative AI model in conjunction with ORCA's photonic hardware, which links quantum devices with traditional processors, and then physically manufactured the peptides to check whether the predictions held up.
The project was cobbled together on weekends and leftover budget from other grants. DTU professor Timothy Patrick Jenkins, who led the work, said the team took that approach because "most innovative science is too scary for foundations." His broader lab, which uses AI to discover proteins for immunotherapies, is largely funded by the Novo Nordisk Foundation.
“We needed to really prove it to convince skeptics that our predictions connect to the real world”— Timothy Patrick Jenkins, DTU professor
Key facts
- 01A Technical University of Denmark team ran a generative AI protein-prediction model alongside a printer-sized ORCA Computing quantum machine to design novel peptides.
- 02The hybrid model outperformed its classical counterpart in the lab, with the biggest gains where training data was scarce.
- 03The project was funded on weekends and leftover budgets from other work, with the Novo Nordisk Foundation backing DTU's broader protein research.
- 04ORCA Computing is also running commercial quantum projects with BP on chemistry and Toyota on design efficiency.
- 05DTU next plans to apply the workflow to larger proteins and to synthetic antidotes for snakebite venom.
Jenkins says a persistent problem for his group has been the shortage of genetic data outside Western populations, which makes it harder to design peptides that work in patients in Asia and Africa. The hypothesis behind the ORCA collaboration was that a quantum-augmented model might generate a more diverse set of candidate peptides, particularly on targets where classical training data runs thin — a pattern the team had seen reported in quantum-assisted image generation.
Jenkins was, by his own account, not the natural customer for this pitch. "I was a huge quantum skeptic," he says, having assumed any real application to his work was "decades away." The lab result appears to have changed the calculus, at least enough to justify moving to larger models and bigger molecules.
The caveats are real. Today's quantum machines are still too small to run frontier-scale AI models end to end, meaning a well-tuned classical system could in principle match or beat the hybrid result on many tasks.
“Quantum is still not very powerful, so the level of complexity that we could encode wasn't a normal-sized antibody, which is what we usually work with”— Jonathan Funk, DTU PhD student
DTU PhD student Jonathan Funk noted that the peptides the team could encode were shorter than the full antibodies his group normally works with. Finding a peptide that binds to a target gene is also only one step in vaccine development; a successful binder does not by itself produce a viable drug.
ORCA Computing chief executive Richard Murray framed the result as an early proof point rather than a breakthrough, acknowledging the field's credibility problem head-on. Quantum computing, he said, "has not ever had really clear near-term examples of usefulness," which is why so many industrial customers treat it as speculative.
“has not ever had really clear near-term examples of usefulness”— Richard Murray, ORCA Computing CEO
ORCA is trying to close that gap with a portfolio of applied projects. Alongside DTU, the company is working with oil major BP on chemistry problems and with Toyota on making its design workflows more efficient — commercial engagements that would only make sense if the hardware can produce results a classical system cannot match on cost or quality.
The DTU team plans to push the workflow toward larger proteins and more cutting-edge generative models next. Jenkins is also exploring whether a quantum-augmented generative model can help design synthetic antidotes for snakebite venom, a neglected disease area with little commercial research funding. "We needed this as an easy way to validate that now we actually have a shot at moving the needle substantially," he says.
The commercial story here is narrower than the headline suggests: a small quantum machine helped a generative AI model produce better candidates in a data-poor corner of biology, and the improvements were verified in the lab. That is a meaningful result for quantum vendors like ORCA, which have spent years fielding the question of what their hardware is actually good for today. It also points to where AI-plus-quantum is most likely to earn its keep in the next few years — not by replacing large classical models on well-mined problems, but by extending generative pipelines into rare diseases, understudied populations, and other targets where the training data simply does not exist.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
