At the GTC conference held in Washington DC, NVIDIA officially presented NVQLink, a open system architecture that directly links GPUs with quantum processors (QPU) to raise a new generation of hybrid supercomputersThe proposal seeks to connect both worlds with an ultra-low latency, high-bandwidth link.
The goal is to provide the communications infrastructure that enables execution control, calibration and algorithms error correction in real time, an essential requirement for quantum computing to translate into useful applications. For the European context and the Spanish ecosystem, this opens the door to hybrid projects in materials science, chemistry or energy with greater stability and performance.
A very low latency link to tame qubits
The basic unit of quantum computing, the qubit, is fragile and prone to error; keeping it useful requires frequent calibrations and correction protocols which must be coordinated with classic high-performance systems. NVQLink aims to be that critical channel, guaranteeing response times short enough for the control algorithms to be effective.
According to the company, this link will allow GPUs and QPUs to function as a single unit, enabling quantum-classical continuous interaction work in areas such as molecular simulation or advanced optimization. The idea is to minimize communication overhead, which is key for qubits to maintain their coherence long enough.
Jensen Huang, founder and CEO of NVIDIA, emphasized that scientific supercomputers will evolve towards tightly coupled hybrid designs with QPUs, where NVQLink will act as a universal bridge between the quantum and the classicalintegrating them into a coherent and scalable system.
Ecosystem and partners involved
NVQLink was developed in collaboration with nine U.S. Department of Energy national supercomputing centersThese include Brookhaven, Fermi, Lawrence Berkeley, Los Alamos, and Oak Ridge. In parallel, architecture is adopting an open strategy and is now compatible with 17 QPU manufacturers y five control system providers.
- QPU manufacturers present in the ecosystem: Alice & Bob, Atom Computing, IonQ, IQM Quantum Computers, Pasqal, Quantinuum, Rigetti, among others.
- Quantum control providers: Keysight Technologies, Quantum Machines, Zurich Instruments, and more technical partners.
The proposal seeks a solution unified and ready to integrate The different approaches to quantum hardware with GPU-based supercomputing reduce the complexity of scaling equipment and standardize communication between platforms.
CUDA-Q as an access route for developers
The entry point to NVQLink will be the software platform NVIDIA CUDA-QFrom this platform, researchers and developers can create and test applications that seamlessly utilize CPUs, GPUs, and QPUs. The goal is to facilitate Hybrid algorithms and workflows that prepare the industry for the quantum-classical supercomputers of the future.
For Europe and Spain, the compatibility with actors such as IQM (Finland), Pasqal (France) or Quantinuum (United Kingdom) This is especially relevant, since It brings technology closer to European R&D programs and could align with high-performance initiatives in university and research centers, provided that agreements and deployments are finalized.
Potential impact in Europe and Spain
Fields like the computational chemistryMaterials science or energy would benefit from stable hybrid computation to simulate complex interactions or explore new compounds. The ability to coordinate QPUs with GPUs in real time is an important step towards high-impact applications in European industrial and academic ecosystems.
From an organizational standpoint, an open architecture like NVQLink promotes interoperability and can accelerate adoption processes in public-private consortia, always subject to the maturity of quantum hardware and the investment priorities of each country.
What is known and what still needs to be determined
Currently, NVIDIA reports compatibility with 17 QPU manufacturers and five controller manufacturers, as well as collaboration with nine DOE national laboratories. However, no public deployment schedules have been detailed beyond the CUDA-Q integration, so there's still a way to go before we know more. Dates, specific configurations, and initial installations.
Although challenges remain in error correction and qubit scaling, the proposal focuses on reduce integration friction between quantum hardware and classical supercomputing, an indispensable element to advance towards practical use cases.
NVQLink focuses on a hybrid model in which GPU intelligence and qubit sensitivity Working side by side; with the support of public laboratories and a diverse group of manufacturers, Europe and Spain are better positioned to take advantage of the first useful applications when quantum hardware reaches the necessary maturity.
