Paper: Simulating 50 qubits
I’m not sure exactly where simulation of qubits will get us going forward, but it is interesting. I recently reviewed the paper “Universal Quantum Simulation of 50 Qubits..."
De Raedt, H., Kraus, J., Herten, A., Mehta, V., Bode, M., Hrywniak, M., ... & Lippert, T. (2025). Universal Quantum Simulation of 50 Qubits on Europe’s First Exascale Supercomputer Harnessing Its Heterogeneous CPU-GPU Architecture. arXiv preprint arXiv:2511.03359. https://arxiv.org/pdf/2511.03359
I’m not sure exactly where simulation of qubits will get us going forward, but it is interesting. I recently reviewed the paper “Universal Quantum Simulation of 50 Qubits on Europe’s First Exascale Supercomputer Harnessing Its Heterogeneous CPU–GPU Architecture” by De Raedt et al.. The authors present the simulator JUQCS-50, which runs on the JUPITER system and uses an array of 16 384 GH200 superchips to simulate a full 50-qubit universal quantum computer for the first time. Their innovations include adaptive byte-encoding of the state vector to reduce memory by a factor of eight, efficient use of LPDDR5 + HBM3 heterogeneous memory, and an on-the-fly network optimization to reduce inter-node communication. For those of us tracing fault-tolerant quantum computing thresholds, this work underscores that classical simulation of universal quantum circuits is still advancing rapidly, tightening the gap to near-term quantum hardware.
Things to consider:
While 50 qubits is impressive, the method still uses byte-encoded state vectors, meaning precision trade-offs may exist in less controlled circuits.
The use of a top-tier exascale supercomputer (JUPITER) raises questions of cost and accessibility for broader quantum algorithm benchmarking.
The near-constant weak-scaling of compute time suggests that future bottlenecks will centre on memory capacity and interconnect bandwidth rather than raw FLOPs.
From a software-engineering perspective the adaptation of MPI + CUDA + unified memory is non-trivial and suggests large-scale simulation remains a niche rather than a mainstream tool.
This simulation milestone does not yet displace the role of real quantum hardware, but it shifts the horizon for when hardware overtakes the best classical simulators.