NVIDIA NVQLink To Connect Quantum Processors With GPUs
NVIDIA NVQLink Unites Classical Supercomputing with QPUs for Quantum-GPU Era
GPU-NVQLink
NVIDIA NVQLink, an open system architecture, was introduced on October 28, 2025, to tightly integrate GPU processing with quantum processors. NVQLink, a rapid link, connects quantum devices to CPU and GPU-based computers for the first time. This innovative event to speed quantum supercomputers was announced at NVIDIA's GTC conference in Washington, D.C.
NVIDIA founder and CEO Jensen Huang calls NVQLink the “Rosetta Stone connecting quantum and classical supercomputers,” launching the quantum-GPU computing era. Huang expects all NVIDIA GPU scientific supercomputers to be hybrid quantum processor-based machines with increased computing power.
Overcoming Quantum's Scaling Roadblock: High-Speed Control
NVQLink is needed because scaling quantum devices is challenging. Qubits, which enable quantum computers to process data differently, are delicate and error-prone. Complex control algorithms, calibration, and advanced quantum error correction (QEC) are needed for quantum systems to work and have applications.
The “symbiotic relationship” between classical and quantum computing was explained by NVIDIA's engineering, semiconductor, and quantum business general manager Tim Costa. Even though quantum computers can solve problems classical systems cannot, optimal control, calibration, and error correction are computationally difficult to implement. Costa stressed that a quantum computer would not be practicable without a big AI supercomputer to tackle control concerns.
For this control process, large volumes of AI computation must be sent via a demanding low-latency, high-throughput connection to the traditional supercomputer, especially as quantum processors grow. The complexity of quantum signals has made it difficult to connect these systems, hindering quantum advances. NVQLink engineers created this important connection environment.
Architect Integration and Performance
NVQLink provides a standardised, turnkey answer to quantum researchers' major integration issues as they add gear. The rapid link lets scientists and engineers call GPU calculations from a quantum processor.
The NVIDIA CUDA-Q software platform lets developers create and test hybrid applications that combine CPUs, GPUs, and quantum processors. QEC requires extensible libraries, which CUDA-Q supports. The link supports trapped-ion, photonic, neutral atoms, and superconducting QPU modalities and is interoperable.
Specifications emphasise real-time capacity for comprehensive error correction:
The GPU-QPU latency is as low as 4.0 microseconds (FPGA to GPU to FPGA).
Maximum GPU-QPU throughput: 400 Gb/s.
With sparsity, 40 PFLOPS (FP4) is leading-edge AI.
The platform accelerates quantum process from calibration to fault tolerance. NVQLink speeds up these crucial tasks:
Real-time QPU calibration requires strong computational coupling to eliminate QPU downtime and improve quantum operations.
Speeding up QEC decoding with low latency and high-throughput processing turns a noisy QPU into a logical, functional one.
Advanced QEC protocols require logical orchestration to execute complicated logical programs using dynamic routing and just-in-time compilation.
Collaboration for Fault-Tolerant Computing Researchers from major supercomputing centres at nine U.S. government laboratories led NVQLink publicly and cooperatively. The DOE's Oak Ridge, Pacific Northwest, Sandia, Los Alamos, MIT Lincoln, Fermi, Berkeley, and Brookhaven labs are participating.
Five control system makers and 17 QPU builders are among the design's hardware partners. QPU partners include IQM Quantum Computers, Alice & Bob, IonQ, Pasqal, Quantinuum, QuEra, and Rigetti. Zurich Instruments, Qblox, QubiC, Quantum Machines, and Keysight Technologies collaborate on controls.
Government officials consider this collaboration strategic. “It must build the bridge to the next era of computing: accelerated quantum supercomputing” to “maintain America’s leadership in high-performance computing,” said Energy Secretary Chris Wright.
He noted that NVQLink provides the technology needed to connect top-tier GPU supercomputers with cutting-edge quantum processors, and that government labs, startups, and industry partners like NVIDIA must work together.
Many partners mentioned NVQLink as a key component for future scalability when they quickly accepted. Pasqal noted that integrating with NVQLink will accelerate their roadmaps to Fault-Tolerant Quantum Computing (FTQC) and utility-scale quantum computing by improving logical architectures.
Alice and Bob agreed that NVQLink addresses crucial FTQC stack layers, indicating that fault-tolerant quantum computers are nearing commercial maturity. Quantum Machines also highlighted how the open architecture increases their microsecond-latency control solution for real-time data exchange and deterministic feedback.
