ORNL, NVIDIA, and HPE move hybrid quantum-AI-HPC from concept to practice
The Department of Energy's Oak Ridge National Laboratory is teaming up with NVIDIA and HPE to push hybrid quantum-classical computing into everyday scientific use. The focus: connect quantum processors with GPU supercomputers, bring error correction research into reach, and test workflows that blend quantum, AI, and HPC for real studies.
U.S. Secretary of Energy Chris Wright framed the goal simply: maintain U.S. leadership in high-performance computing by building a bridge to accelerated quantum supercomputing. He highlighted NVIDIA NVQLink as the fabric that ties GPU systems to emerging quantum processors so scientists can run integrated workloads.
What's being built at OLCF
At Oak Ridge Leadership Computing Facility, teams will use NVIDIA NVQLink and the NVIDIA CUDA-Q software stack to coordinate classical clusters with quantum processors. CUDA-Q also supports large-scale, GPU-accelerated quantum emulation for side-by-side comparisons with hardware.
The testbed will run on an HPE-built NVIDIA GB200 NVL72 system slated for installation in early 2026 at the OLCF data center, which also hosts Frontier, the first exascale supercomputer.
Why it matters for researchers
- Hybrid workflows: Orchestrate GPU/CPU nodes with quantum devices for low-latency control loops, including quantum error correction and variational algorithms.
- Emulation at scale: Use CUDA-Q to simulate qubits on GPUs, mirror hardware behavior, and iterate without device noise or queue limits.
- Noise modeling and AI-assisted decoding: Compare emulator and hardware results, inject realistic noise on the emulator, and train models to mitigate observed errors.
- Protocol-agnostic research: Explore neutral atoms, trapped ions, superconductors, and more-without locking into one hardware path. Interoperate with Frontier and ORNL's Quantum Brilliance and IQM systems.
- Systems research: Develop iterative, low-latency decoders and reduce bottlenecks across the hybrid stack-from device I/O to scheduling and data movement.
Context: where quantum stands
Quantum computing uses qubits, which can occupy multiple states at once through superposition. That property could enable faster routes to solutions in physics, chemistry, and materials science that strain classical methods-even at exascale and for emerging AI models.
The blocker remains error rates. Qubits are fragile, and the winning hardware approach is still undecided. By pairing classical accelerators with quantum devices and high-fidelity emulators, ORNL aims to study error correction strategies and close the loop between theory, simulation, and experiment.
Early testing and validation
Initial testing is underway at a secure HPE factory with direct liquid cooling, where ORNL and NVIDIA teams are running early workloads to validate system performance and stability. These runs will guide how to scale decoders, tune CUDA-Q pipelines, and remove bottlenecks in hybrid workflows.
A track record of scaled collaboration
ORNL and NVIDIA launched Titan in 2009, the first major GPU-CPU hybrid supercomputer. They later delivered Summit with NVIDIA V100 Tensor Core GPUs in collaboration with IBM. Work with HPE led to Frontier, the first computer to cross the exascale threshold in 2022. The OLCF operates as a DOE Office of Science user facility.
Timeline and what to expect
- Now: Early workloads at HPE for stability and performance validation.
- 2025: Expanded hybrid experiments aligned with the International Year of Quantum Science and Technology.
- Early 2026: Installation of the HPE-built NVIDIA GB200 NVL72 system at OLCF and broader user access thereafter.
Big picture
ORNL is investing in quantum research throughout 2025 while maintaining open paths across hardware approaches. UT-Battelle manages ORNL for the DOE Office of Science, which supports foundational research across U.S. labs and universities.
Learn more about the DOE Office of Science
Explore Frontier at OLCF
Your membership also unlocks:
