On November 16, 2026, at the SC26 supercomputing conference in Chicago, the U.S. Department of Energy’s Fermi National Accelerator Laboratory, IBM, and NVIDIA announced a collaborative initiative to establish a modular quantum-classical co-design platform. The joint project co-locates NVIDIA DGX B200 accelerated computing systems alongside advanced IBM quantum processors at Fermilab’s Superconducting Quantum Materials and Systems (SQMS) Center, creating an integrated supercomputing environment specifically tailored for high-energy and nuclear physics simulations.
Integration of Classical Supercomputing and Modular Quantum Hardware
The collaboration addresses an operational bottleneck in scientific computation: bridging the architectural gap between heterogeneous high-performance computing (HPC) clusters and physical quantum processing units (QPUs). By coupling NVIDIA DGX B200 systems with IBM quantum processors—including the IBM Quantum Nighthawk system—the initiative establishes an active testbed for quantum-classical co-design. Fermilab will coordinate application development and researcher access through its HEPCloud distributed computing infrastructure, connecting particle physicists across national laboratories and universities to hybrid workflows.
A core priority of the platform is exploring processor modularity. Because single monolithic quantum processors face physical scaling limits, linking multiple QPUs via low-latency quantum interconnects is essential for scaling computational capacity. In its initial deployment phase, researchers will model and emulate inter-module communication, analyzing how network latency, quantum link errors, and classical feedback loops influence distributed quantum calculations. In subsequent phases, Fermilab intends to integrate its nine-cell superconducting radiofrequency (SRF) cavity hardware into the platform, evaluating these ultracoherent structures as quantum memories, entanglement buffers, and resources for bosonic quantum error correction.
Architectural Specifications and Software Ecosystem
The platform unites specialized hardware layers with newly expanded software interfaces that integrate quantum resources into standard supercomputing schedulers. Central to the software pipeline is the open-source Quantum Resource Management Interface (QRMI), developed jointly by IBM, NVIDIA, Pasqal, and the open-source computing community. QRMI connects directly to the Hierarchical Resources (HRES) plugin architecture in Slurm 26.11, enabling high-performance computing centers to schedule QPUs as shared cluster resources alongside CPUs and GPUs rather than managing them through isolated external queues.
Complementing hardware co-location, NVIDIA introduced the TCBP-JOBS decoder as part of its CUDA-Q QEC 0.9.0 library. The GPU-accelerated decoder combines trajectory-calibrated belief propagation with jittered ordered-statistics decoding (OSD) beam search, enabling real-time error identification across surface, color, and bivariate-bicycle quantum error-correcting codes. In benchmark tests on bivariate-bicycle codes, the decoder reduced offline research calculations from 12 weeks to 45 minutes, delivering up to a 13,000-fold execution acceleration over conventional unsparsified decoding baselines.
| Platform Component | Primary Provider | Key Technical Specifications | Operational Role in Co-Design |
|---|---|---|---|
| Quantum Processing Units | IBM | IBM Quantum Nighthawk, 120 qubits, up to 7,500 two-qubit gates, 100,000 samples/sec | Executes quantum circuits for lattice gauge theories and state preparation |
| Accelerated Classical Compute | NVIDIA | NVIDIA DGX B200 systems (Blackwell architecture GPUs) | Performs tensor-network simulations, real-time control, and error decoding |
| QEC Decoder Architecture | NVIDIA Research | TCBP-JOBS (CUDA-Q QEC 0.9.0), trajectory-calibrated belief propagation | Accelerates syndrome decoding up to 13,000x for candidate error-correcting codes |
| Coherence and Memory Systems | Fermilab / SQMS | Nine-cell ultracoherent niobium SRF cavities, multimode bosonic memories | Serves as quantum memory buffers and substrates for bosonic error correction |
| Workload Orchestration | Open Consortium / Slurm | Slurm 26.11 HRES with open-source QRMI and QDMI interface support | Enables multi-tenant scheduling and allocation of QPUs in HPC clusters |
| Workflow Access Layer | Fermilab | HEPCloud distributed scientific computing portal | Coordinates researcher authentication, job dispatching, and grid federation |
Institutional Background and Precursor Demonstrations
The initiative builds on foundational research conducted under the Superconducting Quantum Materials and Systems (SQMS) Center, one of five national quantum information science research centers established under the National Quantum Initiative Act. Led by Fermilab and encompassing more than 40 partner institutions across academia, government, and industry, SQMS has spent six years advancing superconducting radiofrequency cavities to achieve millisecond-scale photon coherence times.
The collaboration follows successful preliminary demonstrations of hybrid quantum-classical workloads. In earlier benchmarking, Fermilab used NVIDIA CUDA-Q logical layer tools to compress quantum circuit simulation tasks from five months down to three weeks. In parallel computational chemistry experiments with Japan’s RIKEN institute supported by NEDO, IBM quantum hardware coupled with RIKEN’s GPU-accelerated ROQUO system simulated electron interactions in iron-sulfur clusters (Fe4S4) with 79 times higher accuracy than earlier runs, approaching the precision of classical density-matrix renormalization group (DMRG) baselines.
The platform also complements an active co-design project between IBM, Oak Ridge National Laboratory, and AMD, reinforcing federal research efforts under the DOE Advanced Scientific Computing Research program to create an integrated ecosystem of fault-tolerant quantum computing testbeds.
Stakeholder Perspectives from National Labs and Industry
Leaders across the partner organizations emphasized the importance of user-driven co-design in preparing high-energy physics for fault-tolerant quantum architectures.
“This intended platform presents an extraordinary opportunity to bring together IBM’s leadership in quantum processors and networking, NVIDIA’s breakthroughs in accelerated computing and superintelligence, and Fermilab’s scientific community and SQMS expertise in ultracoherent quantum hardware,” said Anna Grassellino, Fermilab chief technology officer and director of the SQMS Center. “By bringing high-energy and nuclear physicists into the process, we can co-design the applications, algorithms and modular architectures needed to make fault-tolerant quantum computing useful for science. This would be an important step toward the capabilities and community needed for a future DOE quantum computing user facility.”
“Scaling quantum and classical together will open new frontiers in high-energy and nuclear physics and beyond,” stated Jerry Chow, IBM Fellow and chief technology officer of Quantum-Centric Computing. “With Fermilab and NVIDIA, we’re exploring how these real-world problems can help us shape the modular, interconnected architectures that will define the next generation of quantum computers and accelerated classical computing.”
“Integrating quantum computing with supercomputing workflows has the potential to dramatically expand how scientists can explore nature’s most fundamental questions,” added Krysta Svore, vice president of applied research for quantum computing at NVIDIA. “Fermilab, IBM and NVIDIA are showing how quantum can be brought into the heart of scientific research with hybrid quantum-classical systems—making quantum processors into powerful new instruments of discovery.”
Significance for Universities, Research Offices, and Facility Planners
The launch of a dedicated modular quantum-classical co-design platform has direct strategic and administrative implications for academic institutions, computing facilities, and university research leadership:
- Elimination of Bespoke Operational Queues: By adopting vendor-neutral scheduling interfaces through QRMI and Slurm 26.11, academic supercomputing centers can integrate on-premises or cloud-hosted QPUs without developing custom job dispatchers. Research computing administrators can allocate QPU processing time using standard project account codes, fair-share scheduling trees, and walltime limits familiar to HPC users.
- Expansion of Research Software Engineering Demands: Developing hybrid algorithms that interleave classical GPU decoders with quantum execution loops requires institutional support from specialized staff. Universities expanding into quantum information science must support the research software engineer role to help domain scientists refactor legacy code for frameworks such as CUDA-Q and Qiskit.
- Alignment with Federal Infrastructure Grants: Institutional researchers targeting Department of Energy and National Science Foundation solicitations can leverage the platform’s benchmarks in proposals. Research offices preparing submissions can consult CASRAI guidance on research infrastructure funding mechanisms to align institutional co-design requests with federal facilities programs.
- Rigorous Workflow Governance and Reproducibility: Hybrid quantum-classical experiments introduce complex software dependencies, including compiler versions, quantum calibration data, and hardware noise characteristics. Research teams must follow systematic guidelines for reproducibility infrastructure and containerized workflows to ensure quantum execution traces can be audited across heterogeneous nodes.
- Compliance with Federal AI and Compute Governance: As classical computing nodes incorporate machine-learning-driven optimization and decoding algorithms, campus compliance officers must verify project adherence to emerging federal standards, such as those outlined in the DOE AI policy framework for secure scientific workflows.
Development Timeline and Subsequent Milestones
The partners outlined a staged rollout for the co-design platform and its associated software components. General availability of Slurm 26.11 with native HRES quantum scheduling support is scheduled for late November 2026, accompanied by reference configurations for the QRMI and QDMI interfaces.
Initial engineering trials at Fermilab will focus on benchmarking communication latency, data throughput, and decoding overhead between the co-located NVIDIA DGX B200 systems and IBM Quantum Nighthawk hardware. Following baseline validation of the interconnect emulation models, the SQMS Center plans to integrate its custom superconducting radiofrequency cavities into the signal chain, establishing proof-of-concept tests for bosonic memory storage and modular quantum networking in scientific research environments.
Sources
- Fermilab: Fermilab, IBM and NVIDIA plan modular quantum computing co-design platform to advance scientific discovery, checked 2026-11-16
- IBM Newsroom: IBM, Fermilab, and NVIDIA Announce Plans for New Co-Design Platform to Accelerate the Development of Next-Generation, Modular Quantum Computers, checked 2026-11-16
- NVIDIA Blog: NVIDIA Expands Slurm to Bring Quantum Computing Into Supercomputing, checked 2026-11-16
- The Next Web: Nvidia teams up with IBM and Fermilab on a quantum codesign platform, checked 2026-11-16
- Unite.AI: NVIDIA, IBM and Fermilab Plan Modular Quantum Computing Platform, checked 2026-11-16







