What would it take for you to trust a quantum computer’s answer about a new drug molecule? For years the honest answer was “not much,” because the machines were too noisy to get chemistry right. A new demonstration from Qedma Quantum Computing and a research group called HQC2 just moved that answer. The team made a quantum chemistry calculation fifty times more accurate, and the gain did not come from a bigger or newer quantum computer. It came from smarter software running on hardware that already exists.
The result makes a fresh, concrete case for quantum error mitigation, the technique that lets today’s noisy quantum processors do real scientific work years before fully fault-tolerant machines arrive.
Qedma and HQC2, a consortium linking the University of Copenhagen, the Technical University of Denmark, and the University of Southern Denmark, ran the test on IBM’s Aachen quantum processor, as detailed in a technical writeup from Quantum Computing Report. The team mapped a water molecule’s potential energy surface onto an eight-qubit register and computed it first with no correction at all. That raw result missed the true energy by roughly 500 millihartree, far too large for any chemist to trust. Layering Qedma’s QESEM software onto the same circuit brought the error down to about 1.5 millihartree, comfortably inside the tolerance chemists call chemical accuracy.
How Qedma’s Quantum Error Mitigation Actually Works
QESEM stands for Quantum Error Suppression and Error Mitigation, and it tackles noise differently than the error correction schemes hardware makers are racing to build. Error correction adds extra physical qubits to protect each logical qubit, which works well but costs a lot on today’s small processors. Error mitigation instead studies exactly how a specific chip misbehaves, then uses that noise map to strip errors back out of the final answer through statistical cancellation.
- Qedma’s software profiles the fractional rotation gates that give chemistry circuits their precision.
- It then applies probabilistic error cancellation to undo what the noise did.
- None of it requires new hardware. It runs today on an IBM system that customers can already rent.
Why Accuracy Matters More Than Qubit Count Here
Chemists have pointed to molecular simulation for years as one of the first places quantum computers could beat classical ones, since molecules are themselves quantum systems that classical machines struggle to model at scale. That promise only holds if researchers can trust the answers. Prof. Stephan P. A. Sauer of the University of Copenhagen, who worked on the study, put it simply: even small errors can throw off predictions about how a molecule behaves.
Picture James, a 58-year-old grandfather waiting on a new treatment for a chronic condition. He has never heard of a water molecule simulation running in Denmark, but it affects him more than he knows. A drug candidate that looks promising in a flawed simulation can waste years and millions of dollars before failing in a real lab, delaying treatments for patients like James. A result that is close but misses chemical accuracy is not much use to a researcher screening a drug candidate on his behalf.
Dr. Asif Sinay, Qedma’s CEO and co-founder, called the demonstration a strong proof of concept for running real chemistry problems on today’s hardware, instead of waiting years for a future generation of fault-tolerant machines.
Part of a Broader Push to Make Near-Term Hardware Useful
This result lands alongside other efforts to squeeze more reliable performance out of today’s imperfect quantum processors. Trapped-ion maker Quantinuum recently cut its own error correction overhead by 3.5 times with a new architecture called Helix, while MIT researchers have been redesigning superconducting qubits to raise operating fidelity. IBM’s own Nighthawk r2 chip is chasing similar reliability gains at greater circuit depth. Qedma is attacking the same problem from the software side rather than the chip side, so the fix can, in principle, run on hardware a customer already owns.
Funding for the work, which includes support from Denmark’s Novo Nordisk Foundation, hints at where this accuracy gain could pay off first. Pharmaceutical and materials researchers need precise molecular predictions, and a modest accuracy boost can change which compounds are worth testing in a lab.
If a fifty-fold accuracy gain can come from software alone, on hardware that already exists, what happens once fully fault-tolerant machines actually arrive? For patients like James, that question might matter more than any qubit-count headline.