PsiQuantum links Construct to Nvidia CUDA-Q for faster sims

PsiQuantum has integrated Nvidia's CUDA-Q platform into Construct, its software suite for fault-tolerant quantum application development. The combination gives developers access to GPU-accelerated simulation for large-scale quantum algorithms, with performance gains that PsiQuantum estimates at up to 450 times faster than CPU-based simulation.

The integration focuses on state-vector simulation, a classical computing method used to model the behaviour of quantum circuits. Simulation remains central to quantum software development, since most useful workloads will require quantum computers far larger than those available today.

Construct users can request early access to CUDA-Q, a GPU-accelerated simulation, within existing workflows. The approach keeps code unchanged while shifting simulation work onto GPUs. Users can also tune acceleration to match their requirements, with reported speedups ranging from about eight times to the top-end figure on multi-GPU systems.

Why It Matters

Validating large-scale algorithms has become a practical constraint in quantum computing. Developers use simulators to check correctness, study error sensitivity, and estimate resource needs. As circuits grow, classical simulation becomes more expensive, with memory and compute demands rising rapidly as qubit counts increase.

PsiQuantum positions Construct as a toolkit for fault-tolerant quantum computing, where error correction and large circuit sizes are expected to be standard. That focus shifts attention from small demonstrations to workflows closer to production software engineering, with repeatable testing and performance measurement.

"As the industry moves toward utility-scale, fault-tolerant quantum computing, rigorous verification of large-scale quantum applications is essential," said Sam Pallister, VP of Quantum Applications at PsiQuantum.

"By integrating NVIDIA CUDA-Q's GPU-acceleration into Construct, we are giving developers the ability to simulate and stress-test algorithms at an advanced scale, mitigating deployment risk," Pallister said.

How It Works

CUDA-Q is Nvidia's platform for hybrid quantum-classical computing. It combines tools for building quantum algorithms with integration into GPU-based classical workloads. According to PsiQuantum, the Construct integration uses Nvidia's cuQuantum simulation engine for GPU-based state-vector simulation.

Hybrid quantum-classical workflows are common in near-term quantum research and are expected to remain relevant in fault-tolerant settings. Many proposed algorithms include classical optimisation loops, as well as pre- and post-processing that runs alongside quantum circuit execution.

Tim Costa, vice president and general manager for Quantum at Nvidia, described the target environment as one that mixes quantum and classical workloads. "Simulating future quantum workloads means modelling a complex hybrid quantum-classical environment," Costa said.

"By integrating the CUDA-Q platform into their workflow, PsiQuantum is able to run GPU-accelerated simulations at the magnitudes they need to deploy large-scale, fault-tolerant quantum applications," he said.

Product Context

PsiQuantum launched Construct in September 2025 and said the tools are being used by industry and academic users. It also highlighted a web application for its Circuit Designer tool, announced in February 2026. Circuit Designer focuses on creating and sharing quantum circuit diagrams, a common format for communicating the structure of algorithms in research and early development.

PsiQuantum is developing photonic quantum computing, which uses light-based components rather than superconducting circuits or trapped ions. The company says the approach aligns with high-volume semiconductor manufacturing, and points to the use of existing cryogenic infrastructure and architectural flexibility as part of its scaling strategy.

The CUDA-Q integration also reflects a broader trend in the quantum sector. Industry timelines for large, fault-tolerant machines remain uncertain, increasing the value of software tools that can model and test larger algorithms without running on quantum hardware. It also raises the importance of partnerships with established computing suppliers that control key classical infrastructure, including GPUs and simulation libraries.

Early access to the GPU-accelerated simulation features is available by request. PsiQuantum says the integration is intended for Construct users running the largest simulation tasks.