Europe’s first exascale supercomputer, JUPITER in Jülich, achieved the full simulation of a 50-qubit universal quantum computer, surpassing the previous record of 48 qubits. This feat required about 2 petabytes of memory and showcases the growing synergy between high-performance classical computing and quantum research
Breaking the 50-Qubit Barrier: A team at the Jülich Supercomputing Centre in Germany, in collaboration with NVIDIA, has set a new record by classically simulating a 50-qubit quantum computer in full detail. They ran the simulation on JUPITER, Europe’s first exascale supercomputer (operational as of September 2025), leveraging its enormous memory and processing resources. The prior record was a 48-qubit simulation achieved in 2022 on Japan’s K-computer. Simulating 50 qubits is a massive leap – quantum state spaces grow exponentially, so each additional qubit doubles the memory needed. In fact, representing the state of 50 ideal qubits required on the order of 2 petabytes of memory, and synchronizing $2^{50}$ amplitude values (quadrillions of parameters) across thousands of nodes Such a task is far beyond ordinary computers; only a machine like JUPITER, with its cutting-edge GH200 Grace-Hopper superchips (integrated CPU–GPU units) and fast interconnects, could pull this off
Engineering Feats: Achieving this simulation demanded both brute-force power and clever software optimizations. The Jülich team upgraded their in-house simulator (JUQCS) to “JUQCS-50”, introducing a hybrid computing approach: when the quantum state data doesn’t fit in GPU memory, the system seamlessly offloads parts to CPU memory with minimal performance loss They also implemented a custom byte-level compression scheme to reduce memory usage by 8× and a dynamic scheduling algorithm to manage data movement across JUPITER’s 16,384 GH200 superchips These innovations allowed the simulation to run efficiently, executing quantum gate operations that entangled all 50 qubits and mirroring a real processor’s behavior in exacting detail Impressively, JUQCS-50 can simulate quantum algorithms like Variational Quantum Eigensolvers (for chemistry) and QAOA (for optimization) that no current physical quantum computer of 50 qubits could reliably run. The team has integrated this simulator into Jülich’s UNified Infrastructure for Quantum computing (JUNIQ), meaning external researchers and companies can now test quantum algorithms on a virtual 50-qubit device via cloud access
Why Simulate Quantum Computers?: Quantum simulators are an unsung hero of the field. They enable scientists to verify and debug quantum algorithms before real quantum hardware is advanced enough to run them By pushing the simulation frontier to 50 qubits, Jülich is giving a head start to algorithm development targeting the next generation of quantum processors. For instance, researchers can experiment with error-correction schemes, complex entangled states, or hybrid quantum-classical algorithms in a simulated environment that guarantees correct results (aside from numerical rounding). This helps identify which algorithms might show “quantum advantage” and deserve testing on actual quantum chips. It’s also a way to benchmark quantum hardware: if a real 50-qubit device comes online, its outputs on certain tasks can be directly compared against JUQCS-50’s predictions to measure fidelity. In essence, advances in classical simulation drive quantum progress by providing a testing ground and setting performance bars for quantum machines to beat
HPC-Quantum Synergy – A Sign of the Times: The achievement underscores a broader trend: the convergence of supercomputing and quantum computing. NVIDIA’s involvement (providing the GH200 accelerated hardware and collaborating via their Application Lab) highlights that classical HPC companies are actively co-designing systems for quantum workloads. We’re seeing the early days of quantum-accelerated supercomputers and supercomputer-accelerated quantum research. For PostQuantumApps and other software teams, this means the ecosystem is preparing for a hybrid future. Developers might run parts of their computations on quantum simulators or use HPC backends to test their quantum-safe protocols under realistic quantum attack simulations. Moreover, Europe’s investment here (JUPITER and the JUREAP program) reflects strategic intent: ensuring that when quantum computers reach 50+ reliable qubits, the software and use-cases to leverage them will be ready. Companies should note that tools like JUQCS-50 could be used to assess the impact of quantum algorithms on their industry problems (e.g. portfolio optimization, molecular modeling) without waiting for physical qubit hardware. The ability to simulate a “useful” size quantum system today is a valuable window into what the near future holds.
Takeaway: Simulating 50 qubits doesn’t mean we have a full quantum computer yet, but it’s a critical milestone. It shows that classical computing is stretching to supplement quantum development, and the bar for quantum supremacy keeps rising. It also implicitly reminds us that classical cryptanalysis of PQC (or even brute-force attacks) could improve with such powerful classical tools – another reason to stay vigilant and adopt strong post-quantum security. In the race toward quantum advantage, achievements like Jülich’s serve as both benchmark and catalyst, ensuring that when quantum hardware crosses the 50-qubit threshold in reality, we’ll be prepared to make the most of it (and defend against it).