NVIDIA Expands Slurm to Bring Quantum Computing Into Supercomputing

Glass Epoch DeskOriginal digestSource: NVIDIA Newsroom

AI-assisted digest, checked against the original source before publishing. How we work

Slurm's new release lets QPU vendors ship one integration file to every supercomputing site, while NVIDIA expands CUDA-Q at SC26.
NVIDIA Expands Slurm to Bring Quantum Computing Into Supercomputing
Photo: Brecht Corbeel · free stock license

NVIDIA announced an expanded Slurm scheduler that brings quantum processing units into supercomputing centers' existing workload workflows. Quantum deployments at HPC sites have grown nearly sevenfold over five years, with close to 30 centers worldwide now deploying more than 50 quantum processors. Since many of these sites already use Slurm for scientific workloads, the extension targets computational chemists, nuclear physicists and materials scientists who want quantum hardware reachable through tools they already know.

The release builds on prior work with IBM, Pasqal and the broader quantum community. It lets QPU builders expose their hardware to every supercomputing center through a single interface defined by a simple file, rather than writing a fresh integration for each site-hardware pair. The change is already adopted by Alice & Bob, IBM, IQM, Pasqal, Quantinuum and MQSC, and supports the QRMI and QDMI interfaces. For HPC centers, it means adding a new quantum processor to existing scheduling and resource-management workflows without bespoke QPU-specific development.

The Slurm layer slots into NVIDIA's open quantum-GPU stack centered on CUDA-Q, which handles hybrid quantum-classical programs and coordinates access alongside Slurm. NVIDIA used SC26 to expand that stack on several fronts. CUDA-Q QEC targets error-correction researchers prototyping decoders, including a new decoder server; a TCBP-JOBS decoder is claimed at 20,000x faster than the industry standard. Fault-tolerant primitives were also extended so application builders can compile ideas realistically — Fermilab used CUDA-Q for studies in three weeks that would otherwise have taken five months.

Two more CUDA-Q pieces round out the releases: Algorithms adds reusable quantum application components, including data-loading helpers and chemistry and materials tooling, while Pulse gives hardware builders a layer for shaping the control pulses that drive qubits, with GPU acceleration for their generation. Ecosystem adoption is also visible: Quantum Motion integrated its QUICHE chemistry framework with CUDA-Q, and Altera and Oxford Quantum Circuits built an FPGA-based controller that links qubit control and readout to GPUs via NVQLink.

What to watch is whether the single-file integration model actually collapses deployment timelines across the roughly 30 sites now installing quantum hardware, and whether QRMI and QDMI remain the interfaces that converge rather than fragment. Open questions include how the 20,000x decoder speedup and Fermilab's compression of five months into three weeks hold up outside reported cases, and whether wider Slurm-mediated access shifts QPU utilization from vendor demos toward repeatable science workloads.

Key facts

  • • Nearly 30 supercomputing centers are deploying 50-plus quantum processors, up nearly sevenfold in five years.
  • • Slurm's new release is adopted by Alice & Bob, IBM, IQM, Pasqal, Quantinuum and MQSC.
  • • Slurm supports QRMI and QDMI interfaces, enabling one integration file per QPU builder.
  • • A TCBP-JOBS decoder in CUDA-Q QEC is claimed 20,000x faster than the industry standard.
  • • Fermilab completed CUDA-Q studies in three weeks versus an expected five months.

Why it matters

For HPC operators, QPU onboarding shifts from bespoke per-site engineering to dropping a file into existing Slurm scheduling, cutting integration cost as more quantum machines arrive. For builders, reusable QRMI/QDMI integrations plus CUDA-Q's error-correction, primitives, algorithms and pulse layers lower the barrier to shipping hybrid quantum-classical workloads — but real-world validation of th

Spotted an error? Send a correction.

Get the day's AI news in one email

The Daily Epoch · 60-second read · unsubscribe anytime

Related

NVIDIA Expands Slurm to Bring Quantum Computing Into… | Glass Epoch