IonQ plans to install its Superion 256 quantum computer at Nvidia's Accelerated Quantum Research Center in 2027, where the company says it will become the facility's first on-premises quantum processing unit.
The system is expected to connect directly to an Nvidia GB200 NVL72 through NVQLink, with CUDA-Q coordinating workloads across the quantum processor and GPU infrastructure. The setup gives Nvidia and IonQ a defined architecture for testing hybrid quantum-GPU computing, but there are no performance results from that configuration yet.
IonQ announced the NVAQC deployment on Sept. 23. The planned research includes portfolio optimization and risk modeling, materials science, computational chemistry, and broader work on quantum-GPU software.
How IonQ and Nvidia plan to connect quantum and GPU computing
Nvidia built NVAQC around the idea that quantum processors will work alongside classical supercomputers rather than replace them. The Boston facility uses GB200 NVL72 systems and CUDA-Q to support hybrid algorithms, quantum-system simulation, and GPU-assisted quantum error correction.
Superion 256 adds trapped-ion hardware to that stack. eWeek previously examined the Superion 256 platform after IonQ unveiled it on Sept. 8, including the company's move toward semiconductor manufacturing and planned 2027 customer deliveries.
IonQ says Superion uses Electronic Qubit Control to replace more complex laser-based controls with electronics integrated onto the chip. The company has fabricated 256-qubit processors and trapped ions in prototype systems, but it has not published full-system workload results for a delivered Superion 256.
That distinction matters because quantum architectures have different control, cooling, connectivity, and scaling requirements. Connecting a QPU to GPU infrastructure is only one part of making a hybrid system useful.
IonQ's recent research offers some early reference points. A study with Synopsys reported 5.9% to 14.6% reductions in total simulation runtime, with physical execution validated on IonQ's 36-qubit Forte system. Another project with Oak Ridge National Laboratory, Nvidia, and the University of Tennessee used generative AI to produce quantum optimization circuits, but every circuit in that study was simulated on an Nvidia H200 rather than executed on quantum hardware.
What eWeek found: the architecture is defined, but the performance case is not
Based on the systems and research disclosed so far, three areas will determine how much the NVAQC deployment tells enterprises:
- The integration is specific, but not benchmarked. IonQ and Nvidia have named the QPU, GPU system, NVQLink interconnect, and CUDA-Q software layer. They have not disclosed end-to-end latency, throughput, reliability, or a GPU-only baseline for the planned configuration.
- Existing studies do not validate the NVAQC system. The Synopsys work used smaller IonQ hardware, while the optimization study ran entirely on GPU-based simulation. Neither measures a physical Superion 256 connected to a GB200 NVL72.
- Operational data will be more useful than another peak benchmark. Enterprise teams should watch for QPU-to-GPU transfer overhead, sustained reliability, error-management behavior, availability, power requirements, and repeatable results against classical or GPU-only execution.
Other projects are already testing how quantum hardware fits into existing infrastructure. A Diraq deployment in an Equinix data center is taking a different architecture into a shared commercial environment, reflecting the broader push to bring QPUs closer to conventional compute.
If Superion 256 arrives at NVAQC in 2027 as planned, the installation will provide a physical test bed for IonQ and Nvidia's hybrid design. The more important milestone will come when that system produces repeatable results that show where a connected QPU improves on GPU-only computing.
Also read: US funding of up to $300 million is backing three competing quantum architectures as researchers work through fabrication, scaling, control, and infrastructure challenges.