NVIDIA and Phasecraft report 15x speedup in quantum molecular simulation

Phasecraft and NVIDIA have announced a fifteenfold speedup in quantum simulations of complex physical systems, according to a report on dev.to. The result pairs Phasecraft's hardware-adaptive quantum algorithms with NVIDIA's accelerated computing infrastructure, and the companies present it as a meaningful step toward practical, fault-tolerant quantum computing.

Rather than running on a quantum processor, the work emulated quantum workloads on classical hardware. The computations ran on systems built on NVIDIA's Hopper GPU architecture, hosted at the University of Nottingham. Phasecraft contributed its proprietary quantum-enhanced Density Functional Theory (DFT) functionals — DFT being a standard technique in physics and chemistry for calculating the electronic structure of atoms and molecules — while NVIDIA's cuQuantum SDK handled the distribution of quantum workloads across multiple processing units.

Scaling the Variational Quantum Eigensolver

The simulations ranged from 4 to 32 qubits, with the most demanding work concentrated in the 24-to-28-qubit range. At the center of the project is the Variational Quantum Eigensolver (VQE), a hybrid classical-quantum algorithm that searches for a molecule's lowest energy state, information that is central to understanding chemical reactions. Emulating that process on classical hardware requires substantial memory and processing speed, which the report says the Hopper-based systems provided. According to dev.to, the fifteenfold gain held as qubit counts grew, a scale at which many other systems lose efficiency.

A dataset for quantum chemistry

The collaboration also produced a large artifact: a database of more than 3,000 unique molecular emulations spanning 13 molecular systems, described as the largest known dataset of its kind built with the VQE method. The stated purpose is training. Better data improves the accuracy of DFT functional training, which should make future molecular simulations more faithful to real-world physical interactions and give researchers more confidence when deciding which compounds to pursue in the lab.

The Q4Bio program

The research was conducted under the Wellcome Leap Quantum for Bio (Q4Bio) program, which asks how new algorithms can deliver quantum advantages for the global health industry. Drug discovery is the primary motivation: traditional pharmaceutical research relies on slow and expensive trial-and-error, and fast molecular simulation offers a way to model how candidate drugs interact with human cells before any laboratory work begins. The report notes that the same techniques for simulating many-body systems could also apply to materials science, energy storage and environmental research.

Why it matters

The announcement is less about quantum hardware than about the software layer that will eventually run on it. Phasecraft's stated strategy is to build and stress-test applications on the best available classical systems now, so the software is mature when fault-tolerant quantum computers arrive and does not become the bottleneck. A fifteenfold speedup on emulated workloads, sustained toward 32 qubits, suggests that GPU-accelerated emulation is a viable research platform today rather than a stopgap.

The dataset matters independently of the speedup. A corpus of over 3,000 VQE emulations across 13 molecular systems gives other researchers both a benchmark and a training foundation, which could compound improvements in simulation accuracy across the field. For pharmaceutical work in particular, faster screening of candidate compounds changes the economics of early-stage research.

One caveat: the figures come from a single report on dev.to summarizing the companies' announcement, and the benchmark methodology has not been independently verified here.