TL;DR — Key Takeaways
- Nvidia, IBM and Fermilab plan to develop a quantum-classical computing platform to research modular quantum systems and error correction.
- Nvidia also introduced TCBP-JOBS, a GPU-accelerated decoder that the company says achieved up to 13,000 times faster error-correction decoding in testing.
- Slurm 26.11 will add capabilities to help supercomputing centers schedule and manage quantum processors through existing HPC workflows.
Nvidia on Monday announced plans to develop a quantum-classical computing platform with IBM and the Department of Energy’s Fermilab, along with new error-correction software and expanded support for quantum processors in the Slurm supercomputing workload manager.
The company’s announcements, made at the SC26 HPC conference in Chicago, address two major challenges for quantum computing: making error-prone quantum processors more reliable and integrating them with the CPU and GPU infrastructure researchers already use.
The proposed quantum-classical platform will build on a long-time quantum hardware collaboration between Fermilab, IBM and Nvidia. The platform will pair Nvidia DGX B200 GPU servers, each equipped with eight Blackwell GPUs, with IBM quantum computers including those powered by IBM’s 120-qubit Nighthawk processors. The partners plan to use the platform to investigate how to connect multiple quantum processors into larger modular systems, as well as how GPU-accelerated computing can support error correction and other operations across those systems.
Fermilab’s Superconducting Quantum Materials and Systems Center will contribute expertise in quantum hardware and interconnects. The lab also plans to involve high-energy and nuclear physics researchers from institutions worldwide in developing algorithms and benchmarks for applications such as particle interactions, neutrino physics and nuclear simulations. A later phase of the project would incorporate Fermilab’s superconducting cavities, which researchers are investigating as a way to store quantum information. The lab also plans to coordinate researcher access through its HEPCloud computing platform.
Separately, Nvidia introduced TCBP-JOBS, a GPU-accelerated quantum error-correction decoder, available through CUDA-Q QEC 0.9.0, Nvidia’s error-correction software library. The decoder analyzes qubit measurements to identify likely errors so they can be corrected, a critical step in developing fault-tolerant quantum computers. Nvidia reported that TCBP-JOBS accelerated error-correction decoding by as much as 13,000 times compared with a configuration of Google’s Tesseract decoder in tests involving specific quantum error-correction codes.
Nvidia also announced updates to Slurm, the open-source workload manager widely used to schedule jobs on supercomputers. The changes would allow HPC centers to incorporate quantum processors into their existing job queues and resource-management policies, reducing the need for custom software integrations.
The updated Slurm 26.11 will add plugin support for Hierarchical Resources (HRES), a feature for managing shared resources accessible from multiple compute nodes. With the new capabilities, researchers could request quantum processor time through the same scheduling tools used for other supercomputing jobs. Developed with IBM, Pasqal and other contributors, the new capabilities use standard quantum interfaces to connect different vendors’ hardware to existing HPC workflows.
Nvidia’s quantum announcements were part of a larger SC26 rollout that included new Rubin HPC GPUs, a Vera CPU rack-scale platform and the Nemotron for Science suite of tools for developing scientific AI agents. Despite the leadership upheaval following former CEO Jensen Huang’s resignation, Nvidia’s SC26 announcements highlight the company’s work, still very much underway it would seem, on hardware, software and research collaborations in HPC, quantum and AI for science.

