Researchers in Germany have achieved a remarkable computational advance by fully simulating a 50-qubit quantum computer. The feat, unprecedented until then, used JUPITER, Europe’s most recent exascale supercomputer, and surpasses the previous record of 48 qubits, demonstrating the growing capacity of high-performance systems.
JUPITER and the new level of quantum computing
For the first time in history, scientists have managed to simulate a 50-qubit quantum computer, a milestone driven by the power of Europe’s exascale supercomputer.
The research team at the Jülich Supercomputing Center, in collaboration with NVIDIA, has consolidated a significant step in quantum computing. They carried out a complete simulation of a universal quantum computer with 50 qubits, something that became viable thanks to JUPITER. This supercomputer, the first of its exascale type in Europe, was officially opened last September at the Jülich Research Center.
This new mark surpasses the previous record of 48 qubits, also set by Jülich experts in 2019, using the K computer in Japan. In addition to redefining the limit of simulation capacity, the progress highlights JUPITER’s immense potential, promising to accelerate the development of future quantum algorithms and technologies.
The relevance of quantum simulations for the technological future
Quantum computer simulations are essential to boost the advancement of research in this area. Scientists use them to test new algorithms, verify results obtained in experiments and analyze the behavior of future quantum systems, even before the physical hardware reaches the power necessary to perform such operations.
Among the algorithms that arouse great interest from researchers are the Variational Quantum Eigensolver (VQE), crucial for the study of molecules and materials, and the Quantum Approximate Optimization Algorithm (QAOA), developed to solve optimization challenges in segments such as logistics, finance and artificial intelligence.
The exponential challenge of mimicking quantum systems
Recreating the functioning of a quantum computer on a conventional supercomputer is a task of enormous complexity, given the exponential expansion of difficulty with each qubit added. Each new qubit doubles both the memory and processing power required for accurate simulation.
While a typical laptop can process simulations with up to approximately 30 qubits, the 50-qubit simulation represents a giant leap, requiring around 2 petabytes of memory. This volume is equivalent to an impressive two million gigabytes, illustrating the colossal barrier overcome to advance the development of quantum computing.
“Currently, only the largest supercomputers on the planet can deliver such a level of performance,” said Professor Kristel Michielsen, director of the Jülich Supercomputing Center. “This scenario demonstrates the deep connection between the progress of high-performance computing and quantum research today.”
The simulation seeks to replicate in detail the quantum behavior of a real processor. Each operation, such as applying a quantum gate, impacts more than 2 quadrillion complex numerical values — a “2” followed by 15 zeros. These values need to be kept in sync across thousands of computing nodes so that the behavior of the quantum processor is reproduced with maximum precision.
NVIDIA GH200 superchips drive new world record
The technological advancement achieved was significantly dependent on the NVIDIA GH200 Superchips, integrated into the JUPITER system. These chips provide a fluid connection between central processing units (CPUs) and graphics processing units (GPUs), allowing data that exceeds GPU memory to be temporarily stored in CPU memory, without compromising high performance.
To capitalize on this innovative architecture, engineers at the NVIDIA Applications Lab, a collaboration between the Jülich Supercomputing Center (JSC) and NVIDIA, optimized Jülich’s quantum simulation software, known as the Jülich Universal Quantum Computer Simulator (JUQCS). The updated version, named JUQCS-50, performs quantum calculations efficiently even when a portion of the data needs to be transferred to the CPU memory.
Additionally, the research team implemented a byte encoding compression technique, which resulted in a reduction in memory demand by eight times. This has been combined with a dynamic optimization system, which continuously improves the exchange of information between the more than 16,000 GH200 Superchips.
“With the JUQCS-50, we are able to emulate universal quantum computers with high fidelity and explore questions that no existing quantum processor is capable of solving yet,” said Professor Hans De Raedt from the Jülich Supercomputing Center and lead author of the pre-published study.
Expanding access to innovation in quantum research
JUQCS-50 will be made available for use by external research organizations and companies through JUNIQ – the Jülich Unified Infrastructure for Quantum Computing. The researchers’ expectation is that the tool will serve both as a fundamental scientific resource and as a reference point for evaluating future supercomputers.
This project was designed and developed under the JUPITER Research and Early Access Program (JUREAP). “Early collaboration allowed hardware and software to be designed together during the JUPITER construction phase, with close cooperation between Jülich experts and NVIDIA a crucial step towards realizing the full potential of this exascale system,” explained Dr. Andreas Herten, member of the JUPITER project team at Jülich and co-author of the study.
JUPITER’s financing is the result of a joint effort between several entities. Half of the funding comes from the European Partnership for High Performance Computing (EuroHPC JU). A quarter of the funding is provided by the Federal Ministry for Research, Technology and Space (BMFTR, formerly BMBF), and the remaining quarter comes from the Ministry of Culture and Science of the State of North Rhine-Westphalia (MKW NRW), through the Gauss Center for Supercomputing (GCS).
