#perlmuttersupercomputer

NERSC Requests Proposals for 2026: Quantum Information Science Research Can Use 20,000 GPU Node Hours

QIS@Perlmutter

The National Energy Research Scientific Computing Centre (NERSC) has issued the QIS@Perlmutter 2026 Call for Proposals for quantum information science (QIS) experiments on the powerful Perlmutter supercomputer. This contest offers large computing resources with up to 20,000 GPU node hours for accepted proposals. Berkeley Lab hosts NERSC, the U.S. Department of Energy Office of Science's mission computing facility, which supports physics, chemistry, and materials science research.

Boosting QIS and Hybrid Computing

A key goal of QIS@Perlmutter is to develop quantum tools, methodologies, software, and algorithms to further QIS. Hybrid computing, which links quantum systems to classical HPC to promote materials science and chemistry, is a priority.

Accepted quantum science research includes many significant topics. All aspects of quantum information science are project-friendly:

Quantum chemical and material modelling.

Compilation and simulation schemes for quantum circuits.

Quantum computing error prevention.

Software development for quantum computing stacks.

Developing and testing hybrid quantum-classical codes.

Studies comparing HPC with quantum computing systems.

Quantum physics is accelerated by employing Perlmutter's GPUs and other cutting-edge classical computers to solve challenging problems.

High-performance Perlmutter resources and software

This allocation uses the NVIDIA A100 GPU-powered Perlmutter supercomputer. For the NERSC 2026 Allocation Year, selected candidates can use up to 20,000 GPU node hours. Importantly, each Perlmutter GPU node has four A100 GPUs.

Researchers will have access to modern quantum research software and libraries. These include NVIDIA's CUDA-QX and cuQuantum tools, which simplify GPU tensor network and state vector simulation. Qiskit, PennyLane, Quimb, and ITensor are better, more accessible software.

Additionally, winners will work with Nvidia and NERSC staff. This relationship aids recipients with workflow optimisation, project issues, and GPU use.

Qualifications and Allocation Year

All applicants can apply, and NERSC users are not required. Computer science, physics, and materials researchers in academia, industry, and national labs should submit proposals.

In addition to quantum information science research, proposals are evaluated on utilisation criteria:

Leveraging Perlmutter Features: Candidates must demonstrate a well-planned approach to using the system's all-flash scratch file system and over 6,000 NVIDIA A100 GPUs.

Planning for Doudna Capabilities: Proposals should describe how they will leverage the future supercomputer's functionalities. Interactive workflows, Quantum-X800 InfiniBand networking, and NVIDIA Vera-Rubin CPU-GPU compatibility are included.

Social and DOE Impact: Projects that assist the Department of Energy Office of Science's research aims or society will be prioritised.

The allotment spans January 16, 2027, to NERSC 2026. Projects must finish by this date. Project winners must submit a final summary and bimonthly progress reports.

Application and Open Science Needs

Applications are accepted online and reviewed continuously. Please submit by March 1, 2026, to be considered.

Since NERSC is a national user facility of the Department of Energy Office of Science dedicated to open scientific research, all awarded projects must be published or presented in open forums. Every project must fulfil NERSC permissible usage policies. An “Institutional User Agreement” is needed if the study doesn't secure federal funding.

The Perlmutter supercomputer simulates quantum devices in unparalleled detail. The Perlmutter Supercomputer

Effective quantum hardware development has improved because to the Perlmutter supercomputer at the National Energy Research Scientific Computing Centre (NERSC). Perlmutter has been hired to simulate large-scale quantum circuits and quantum microchips in unprecedented detail to increase quantum chip performance.

These conventional simulations aim to provide broad, high-fidelity analysis that accelerates quantum software and hardware development, comparison, and refinement. These extensive classical computations are essential for quantum hardware design and verification, notwithstanding their cost.

A First-Time Physical Simulation of the Microchip

A superconducting quantum microprocessor created by UC Berkeley's Quantum Nanoelectronics Laboratory and Berkeley Lab's Advanced Quantum Testbed (AQT) was simulated in groundbreaking simulation. Simulation of device assembly and functioning at the full-wave physical level was the key goal. Its realism in material, layout, and wiring distinguishes it from other simulations that treat the quantum device as a “black box”. The simulation, which included behaviour over four orders of magnitude, was complicated. Time-domain PDEs for entire device electromagnetic (Maxwell) systems were solved by researchers. Using circuit parts with nonlinear behaviour, the physics model allowed researchers to observe the chip's spectrum and transient responses. This time-domain strategy is crucial because it avoids the transient and nonlinear effects that many frequency-domain or reduced models miss. Simulated object was a 0.3 mm thick, 10 mm square multi-layered chip. To accurately capture electromagnetic wave propagation, the researchers discretised the chip into 11 billion grid cells, including one-micron etchings. To manage this detail, massive processing was needed. The simulation employed over 7,000 of Perlmutter's 7,168 NVIDIA GPUs in a day. The team used the full system to conduct over a million time steps in under seven hours. Researchers tested three circuit designs in one day at this speed. Device-level, time-domain simulations at this scale provide quantum engineers with performance information. It captures nonlinear interactions and realistic noise coupling. This feature considerably improves crosstalk and gate fidelity forecasts compared to simplified or frequency-domain models. By accurately recreating large parts of the hardware, researchers can find hidden resonances, emergent failure modes, and spectral coupling issues that smaller models may miss.

Digging into Quantum Circuits and Algorithms