Qblox links NVIDIA DGX Spark to quantum control with sub-7.2 µs latency
Qblox, a Delft-based maker of quantum control systems, says it has integrated the NVIDIA DGX Spark with a quantum system controller using NVIDIA NVQLink. The company reports measured round-trip latency below 4.0 µs on the NVIDIA RTX PRO 6000 Blackwell and below 7.2 µs on the DGX Spark.
- Qblox integrated NVIDIA DGX Spark with a quantum system controller using NVIDIA NVQLink.
- Reported round-trip latency is below 4.0 µs on the RTX PRO 6000 Blackwell and below 7.2 µs on the DGX Spark.
- GPU-accelerated calibration updates qubit parameters during experiments, and GPU decoders handle error-correction data in real time.
- The same CUDA-Q code can move from a desktop system to data center GPU servers, according to Qblox.
- The work follows NVQLink news from October 2025 and CUDA-Q real-time API support from March 2026.
GPUs inside the real-time control loop
According to Qblox, the Qblox cluster has been integrated with several NVIDIA GPUs. This allows real-time control of hybrid quantum-classical computing. The company says the result is that a single desktop system can host quantum, AI and simulation workloads through NVIDIA CUDA-Q.
Qblox says the same code can then be moved to GPU servers in data centers as systems grow. The integration puts GPU acceleration into the real-time control loop, which the company describes as a key element for scaling quantum systems to support logical qubits.
The Q1 sequence processors from Qblox and LINQ provide the low latency and high throughput the Qblox cluster needs to work together with the GPU, according to the company.
Faster calibration and error correction
Qblox describes two main uses. The first is calibration. GPU-accelerated routines update qubit parameters while experiments are running, instead of between them.
The second is quantum error correction. Syndrome data reaches GPU-based decoders with low latency, which Qblox says makes it possible to identify and correct errors in real time.
The announcement builds on earlier work between Qblox and NVIDIA on NVQLink, announced in October 2025, and on Qblox support for the NVIDIA CUDA-Q real-time API in March 2026. Qblox says this step moves the integration from architecture to measured hardware performance, showing that it can deliver the low-latency performance needed to connect quantum control to real-time GPU computing.
What the companies say
Niels Bultink, CEO and co-founder of Qblox, said: "Les ordinateurs quantiques ne pourront évoluer qu'au rythme des systèmes de contrôle classiques qui les entourent." In English, he argued that quantum computers can only scale as fast as the classical control systems around them, and that GPUs now join this control loop. He said Qblox is proud to work with NVIDIA's NVQLink to bring faster calibration and error correction to the quantum sector.
Sam Stanwyck, director of quantum products at NVIDIA, said DGX Spark puts an AI supercomputer on a quantum researcher's desk. According to him, the Qblox integration with NVQLink and CUDA-Q connects that computing power to quantum hardware and combines simulation, real-time calibration and error correction in one development platform.
What it means for researchers
For research groups, the pitch is a shorter path from desk to data center. Stanwyck said researchers can design and test new approaches on their desktop computer, then adapt the same code to GPU supercomputers as their quantum systems grow.
The latency figures are Qblox's own reported measurements. The source does not give pricing, availability dates or details of the test setup.
