Chinese Researchers Achieve 98% Scalability in neural-network Quantum Chemistry on Sunway Supercomputer

What Chinese researchers achieved this week on the Sunway supercomputer is a breakthrough in neural-network quantum chemistry. They scaled quantum simulations to real molecular sizes, reaching 98% weak scaling efficiency across 37 million processor cores. This advancement bridges AI and quantum science, enabling accurate modeling of complex quantum systems on existing supercomputing resources. The team demonstrated that machine learning can handle large-scale quantum problems, marking a significant step toward practical applications in materials and drug discovery.

Advancing Quantum Chemistry with Classical Supercomputing

Researchers are demonstrating significant progress in simulating quantum chemistry using classical supercomputing resources, a development that could accelerate materials discovery and molecular modeling. A team of Chinese researchers recently leveraged the nation’s new Sunway supercomputer to scale neural network quantum states (NNQS) to realistically sized molecules, bridging the gap between artificial intelligence and quantum science. This achievement highlights the potential for tackling complex chemical simulations with existing computational infrastructure, rather than waiting for fully realized quantum computers.

The project ran on an impressive 37 million processor cores, achieving 92% strong scaling and 98% weak scaling, according to the findings. This level of efficiency indicates a near-perfect synchronization between the algorithm and the supercomputer’s architecture, a rare feat at this scale and essential for performing large quantum simulations. The Sunway supercomputer utilizes SW26010-Pro chips, built with clusters of small compute cores and local memory, enabling precise data control. This architecture proves particularly well-suited to the repetitive training loops inherent in deep learning, which are crucial for NNQS calculations.

Building on this success, the researchers are showing that machine learning can accurately model complex quantum systems, offering a viable pathway for simulating real materials and molecules. The NNQS technique involves training a neural network to represent the possible arrangements and movements of electrons within a molecule, requiring the generation of numerous random samples and the calculation of their “local energy.” Both steps are computationally demanding, yet the Sunway supercomputer’s performance demonstrates the feasibility of this approach with today’s classical hardware. This advancement could significantly reduce the time and resources needed for molecular design and materials discovery.

Scaling Neural Networks for Molecular Simulations

The Chinese researchers achieved remarkable scaling efficiency with their neural network quantum states (NNQS) implementation on the Sunway supercomputer, demonstrating 92% strong scaling and 98% weak scaling across 37 million processor cores. This performance indicates a near-perfect synchronization between the algorithm and the supercomputer’s architecture, a significant hurdle overcome in large-scale quantum simulations. Achieving such high scalability is crucial because the computational demands of accurately modeling molecular systems grow exponentially with their size and complexity, necessitating increasingly powerful computing resources. This breakthrough suggests a viable path towards simulating larger, more realistic molecules than previously possible with classical methods.

The Sunway supercomputer’s architecture, built around the SW26010-Pro chips and their clustered cores with local memory, proved particularly well-suited to the demands of NNQS calculations. Generating the vast numbers of random samples needed to approximate molecular wavefunctions, and subsequently calculating the local energy for each sample, are inherently parallelizable tasks. The system’s fine-grained control over data access minimized bottlenecks, allowing researchers to effectively distribute the workload across the massive core count. According to the team, this efficient parallelization is key to overcoming the “computational brutal[ity]” associated with accurately simulating quantum systems.

Building on this success, the researchers are demonstrating the potential of machine learning to model complex quantum systems with sufficient accuracy for practical applications in materials science and drug discovery. This approach offers a compelling alternative to traditional quantum chemistry methods, which often struggle to scale to systems of realistic size. While not a replacement for future quantum computers, this work highlights the power of leveraging existing classical supercomputing infrastructure to advance our understanding of quantum phenomena. The team’s findings suggest that machine learning-accelerated quantum simulations can become a valuable tool for accelerating materials design and discovery processes.

This achievement by Chinese researchers using the Sunway supercomputer demonstrates a powerful synergy between advanced computing and quantum chemistry. The 98% scalability represents a significant step toward modeling complex molecular systems with unprecedented accuracy, without needing fault-tolerant quantum hardware. For industries relying on materials discovery and drug design, this offers a pathway to accelerate innovation using existing resources.

The implications extend beyond quantum computing to broader scientific modeling, as machine learning efficiently tackles previously intractable simulations. This development could enable researchers to explore and understand complex chemical processes, ultimately informing advancements in diverse fields like energy and medicine, and paving the way for more sophisticated computational methods.

Quantum News

Quantum News

As the Official Quantum Dog (or hound) by role is to dig out the latest nuggets of quantum goodness. There is so much happening right now in the field of technology, whether AI or the march of robots. But Quantum occupies a special space. Quite literally a special space. A Hilbert space infact, haha! Here I try to provide some of the news that might be considered breaking news in the Quantum Computing space.

Latest Posts by Quantum News:

IBM Remembers Lou Gerstner, CEO Who Reshaped Company in the 1990s

IBM Remembers Lou Gerstner, CEO Who Reshaped Company in the 1990s

December 29, 2025
Optical Tweezers Scale to 6,100 Qubits with 99.99% Imaging Survival

Optical Tweezers Scale to 6,100 Qubits with 99.99% Imaging Survival

December 28, 2025
Rosatom & Moscow State University Develop 72-Qubit Quantum Computer Prototype

Rosatom & Moscow State University Develop 72-Qubit Quantum Computer Prototype

December 27, 2025