Quandela Validates Photonic QPU Link to NVIDIA NVQLink, Targeting Real-Time HPC and AI Workloads
Updated
Updated · HPCwire · Jun 23
Quandela Validates Photonic QPU Link to NVIDIA NVQLink, Targeting Real-Time HPC and AI Workloads
3 articles · Updated · HPCwire · Jun 23
Summary
ISC 2026 results showed Quandela experimentally validated a low-latency path linking its photonic QPU with an NVIDIA GPU host and an FPGA-based Quantum System Controller through NVQLink.
The setup is meant to cut the delays of cloud API access and job queues, letting quantum processors operate more like collocated accelerators inside GPU-driven HPC systems.
Photonic quantum machine learning is the first target, including quantum reservoir computing, feature maps and hybrid neural networks, where reused optical configurations and fast sampling make system latency critical.
Quandela said existing HPC schedulers would still handle reservation and accounting, while active GPU-QPU sessions bypass repeated cloud-style orchestration; the work builds on its MerLin framework and MosaiQ platform.
The validation points toward on-premise or dedicated-data-center deployments for HPC centers, sovereign AI and quantum programs, and future NVQLink-enabled MosaiQ systems.
In June 2026, Quandela achieved a major milestone by experimentally validating the integration of its photonic Quantum Processing Units (QPUs) with NVIDIA’s accelerated computing infrastructure using NVQLink. This breakthrough marks a shift from slow, cloud-based quantum access to a tightly integrated, low-latency model, enabling quantum data processing in just 30 milliseconds instead of several seconds. This speed boost is crucial for advancing hybrid quantum-classical high-performance computing and AI workloads. Quandela’s MerLin platform already lets users run models on real quantum devices, showing the immediate benefits of this new, faster integration approach.