At the SC25 high-performance computing conference, NVIDIA presented its vision for integrating next-generation computing architectures, focusing on a specialized interconnect solution designed to bridge classical and quantum processing units.

The company recently expanded its interconnect portfolio to include a protocol enabling high-speed communication between graphics processing units (GPUs) and quantum processing units (QPUs). This technology builds upon NVIDIA’s established GPU interconnect architecture, adapted to meet the anticipated demands of hybrid quantum-classical computing environments. While fully scaled quantum processors are not yet commercially available, several cloud providers—including Amazon Web Services, IonQ, and Rigetti—already offer access to early-stage quantum systems.

Industry analysts suggest that future supercomputers will increasingly incorporate quantum accelerators to tackle complex problems, while quantum processors themselves will depend on classical supercomputers for control and validation. In this model, GPUs are expected to orchestrate quantum hardware operations and manage computationally intensive tasks that exceed current quantum capabilities.

The interconnect is paired with an open programming platform that allows developers to build applications utilizing QPUs, GPUs, and CPUs simultaneously. This approach addresses a fundamental need in quantum computing, where most practical algorithms require substantial classical computing support for functions such as system control, parameter optimization, and error correction.

Quantinuum has been announced as the first quantum hardware provider to adopt the new interconnect, planning integration with its Helios quantum system. NVIDIA states that the combined solution addresses critical challenges in control and error mitigation, enabling the development of hybrid applications. Reported technical specifications include AI performance reaching tens of petaflops at reduced precision, with a data throughput of hundreds of gigabits per second and latency under five microseconds.

Interest in the platform spans leading research institutions worldwide. In the Asia-Pacific region, scientific organizations in Japan, South Korea, Taiwan, Singapore, and Australia are among early adopters. Across Europe and the Middle East, supercomputing and quantum technology centers in Italy, Denmark, Germany, Poland, the Czech Republic, and Saudi Arabia are also engaging with the technology.

In the United States, several national laboratories—including those at Los Alamos, Sandia, Berkeley, Oak Ridge, and partners at the Massachusetts Institute of Technology—are preparing to implement the interconnect framework within their advanced research infrastructures.