Researchers at the Jülich Supercomputing Centre in Germany have completed the first full simulation of a 50-qubit universal quantum computer, a feat that required harnessing every layer of Europe’s most powerful machine to keep pace with the exponential demands of quantum mechanics. Their simulator, called JUQCS-50, ran on JUPITER, the continent’s first exascale supercomputer, which went live after its September 2025 inauguration at the Forschungszentrum Jülich campus in North Rhine-Westphalia.
The results, published in June 2026 in the peer-reviewed journal Future Generation Computer Systems, surpass the previous record of 48 simulated qubits and sharpen a question that hangs over the entire quantum industry: how much longer can classical supercomputers verify what quantum processors claim to do?
Why 50 qubits is an inflection point
A quantum computer with 50 qubits can exist in 2^50 states simultaneously. Simulating that on a classical machine means storing and manipulating roughly 1.13 quadrillion complex numbers. Each additional qubit doubles the memory and compute required, so the jump from 48 to 50 qubits did not add a small increment. It quadrupled the workload compared to the prior record.
That distinction matters because real quantum chips already contain far more physical qubits. IBM’s Condor processor, unveiled in late 2023, has 1,121. Google’s Sycamore chip used 54 qubits in its landmark 2019 quantum supremacy experiment. But physical qubit counts and full classical simulation are different things entirely. Physical qubits are noisy and error-prone; a classical simulation tracks every amplitude with perfect precision. When a classical machine can fully simulate a quantum circuit, it provides a ground-truth check on whether the quantum device actually produced a correct result. Lose that ability, and the field loses its most reliable referee.
At 50 qubits, that referee now needs an entire exascale supercomputer to do its job. At 60 qubits, the memory requirement would balloon by a factor of roughly 1,000. At 70, it would take about a million times more. The classical verification window is not just narrowing. It is approaching a hard wall.
How JUPITER made it possible
JUPITER is built on NVIDIA’s Grace Hopper GH200-class architecture, a hybrid design that pairs energy-efficient ARM-based CPUs with powerful GPUs across thousands of interconnected nodes. The EuroHPC Joint Undertaking, the EU body that co-funded the machine alongside German federal and state partners, rates it at one exaflop: one quintillion floating-point operations per second.
For the JUQCS-50 simulation, that heterogeneous layout proved critical. The 50-qubit state vector had to be distributed across the memory of many nodes simultaneously, with GPUs handling the heavy linear algebra while CPUs managed data movement and orchestration. According to the team’s arXiv preprint, which provides the full technical detail behind the journal paper, the simulator achieved a measured speedup over the 48-qubit benchmark rather than merely matching it. That suggests JUPITER’s architecture scales efficiently for this class of problem, not just brute-forcing it with more hardware.
The team has not disclosed total node-hours consumed, energy costs, or raw performance logs for the 50-qubit run. Those numbers would allow independent researchers to calculate cost-per-gate metrics and compare JUPITER’s efficiency against U.S. exascale systems like Frontier at Oak Ridge National Laboratory. Without them, the speedup claim rests on the peer-review process rather than open replication.
What the result does and does not settle
The simulation does not prove or disprove that any existing quantum computer has achieved genuine quantum advantage. What it does is establish, with peer-reviewed precision, the current boundary of classical capability. If a quantum device running a 50-qubit circuit produces an output, JUPITER can now check that output exhaustively. Beyond roughly 55 qubits, based on the same exponential scaling, no existing or near-term classical machine is likely to manage the same feat.
That boundary has practical consequences. Google’s 2019 Sycamore experiment claimed quantum supremacy at 54 qubits, but subsequent classical algorithms and supercomputer runs chipped away at the claim, showing that clever software could narrow the gap. The Jülich team’s work updates the scoreboard: at 50 qubits with universal circuits, classical simulation is still feasible but only barely, and only on the most powerful hardware on the planet.
Several open questions remain. The published materials do not specify which types of quantum circuits were simulated, whether random circuits of the kind used in supremacy experiments, structured algorithms like Grover’s search, or something else. Circuit depth and gate set matter enormously for simulation difficulty; shallow random circuits are a different challenge than deep, highly entangled ones. The team also has not indicated whether a follow-up at 51 or 52 qubits is planned, though the scaling pattern suggests JUPITER could potentially handle a qubit or two more before hitting its memory ceiling.
Why the classical verification window is closing within a handful of qubits
For governments and companies pouring billions into quantum hardware, this milestone doubles as a countdown clock. Classical verification is the field’s quality-control mechanism. Once quantum processors routinely operate beyond the reach of any supercomputer, the community will need alternative methods to maintain confidence in results. Those alternatives could include cross-checking between quantum devices, cryptographic verification protocols, or partial simulation techniques, but none yet offers the exhaustive certainty of a full classical simulation.
Europe’s investment in JUPITER was driven partly by digital sovereignty goals and a desire to ensure the continent has its own top-tier computational infrastructure rather than depending on U.S. or Asian systems. The 50-qubit simulation gives that investment a concrete scientific payoff and positions the Jülich team at the center of an increasingly urgent international effort.
But the exponential math is unforgiving. The same scaling law that makes quantum computers powerful makes them impossible to fully simulate on classical hardware past a certain point. The Jülich team has drawn that line at 50 qubits more precisely than anyone before. The next few qubits will determine whether classical machines get one last look, or whether the quantum world pulls away for good.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
