Access and program Quantum Processing Units without friction.
Scaleway is your gateway to the leading edge of European quantum hardware. We aggregate a growing catalog of real Quantum Processing Units (QPUs) from major players like Pasqal, Quandela and IQM. Benefit from a fully sovereign ecosystem that connects you directly to the latest photonic, neutral atom, and superconducting technologies.
Accelerate your development with the market's most powerful GPUs, including NVIDIA B300 NVLink clusters, managed and optimized for quantum emulation. Our platform is code-first and fully compatible with open-source SDKs like Qiskit, Cirq, CUDA-Q, and PennyLane. Write your code once, test it at scale on GPU, and deploy it seamlessly to real hardware.
Break free from proprietary lock-in with a truly agnostic platform designed for end-users. We offer total financial transparency with a billing model adapted to your usage: pay per shot, per hour (emulation), or via dedicated booking. You control your budget while exploring the best technology for your specific use case.
Best for local optimization and simulating materials properties
Up to 100 neutral-atom qubits (analogical)
Modalities: Per shot & emulations
Best for general purpose, variational algorithms & education
Up to 54 transmon qubits (gate-based)
Modalities: Per shot & booking
Best for machine learning, sampling & optimization
Up to 12 single-photon qubits
Modalities: Per shot, per booking & emulations
Best for high fidelity algorithms and non-local optimization problems
Up to 12 trapped-ion qubits (gate-based)
Modalities: Per shot & emulations
Open source and GPU-optimized emulator runtime
Up to 38 logical qubits (gate-based)
Modalities: Emulation only
Open source emulation for discovery & prototyping
Up to 36 logical or noisy qubits (gate-based)
Modalities: Emulation only
Open source emulation for discovery & prototyping
Up to 36 logical qubits (gate-based)
Modalities: Emulation only
Qiskit is the most widely used open source SDK in the industry, Qiskit provides clear tools to design and create quantum programs. It is the go-to choice for developers looking for a vast community.
(Soon) Qibo is an open source framework designed for versatility and performance. Qibo distinguishes itself with a hardware-agnostic design that supports various QPUs, the best choice to develop across different architectures.
Cirq is designed specifically on the precise control of quantum circuits and hardware constraints. It is particularly powerful for algorithm developers who need to optimize for specific hardware and handle noise mitigation.
NVIDIA CUDA-Q™ is the QPU-agnostic platform for accelerated quantum supercomputing.
Perceval is the reference toolkit for Photonic quantum computing. Perceval allows you to simulate Linear Optical Quantum Computing (LOQC) by modeling light sources, interferometers, and detectors. It is the native bridge to Quandela’s QPUs, enabling you to encode information into the properties of light with high fidelity. Scaleway is integrated as a first party provider
Pulser is the framework for Neutral Atom quantum computing. Unlike gate-based SDKs, Pulser operates at the pulse level, allowing you to control the physical evolution of the atoms directly. It is indispensable for users targeting Pasqal’s hardware to perform analog quantum simulation and optimize atomic register layouts.
Pennylane is the leading library for differentiable quantum programming. PennyLane integrates seamlessly with classical ML libraries like PyTorch, allowing you to treat quantum circuits as trainable neural network layers. It is the standard for developing variational algorithms (VQE, QAOA) on our infrastructure.
MerLin is a specialized library designed to run deep learning workflows on photonic processors such as Quandela. MerLin optimizes Quantum Neural Networks (QNN) specifically for light-based architectures, offering a direct path to executing efficient, energy-saving AI models on photonic hardware.
Benefit from our Quantum Adoption Program to onboard in quantum
De-risk your innovation. Scaleway supports your discovery as a technological partner of the Quantum Adotpion Program. This program connects industry leaders with quantum experts and provides place to explore hybrid use cases on our sovereign infrastructure.
We back the bold. Quantum is a priority sector for us. Through our Startup Program, we empower early-stage quantum companies with significant cloud credits (up to €36k), visibility, and technical mentorship. Build your quantum-enhanced application, train your QNNs on our Quantum-as-a-Service, and host your data without burning your cash.
First Perceval Quest: When AI meets Quantum
The First Perceval Quest was jointly organized by Quandela and Scaleway to explore the intersection of quantum computing and machine learning through one of the most iconic machine learning benchmarks: the MNIST dataset.
The challenge tackled MNIST problem using hybrid quantum models on a subset of the original dataset. The MNIST dataset consists of 70,000 handwritten digit images, each 28x28 pixels.
This challenge ran from November 2024 to March 2025. Overall, 64 teams joined the Perceval Quest and 11 were selected for the final phase.
Photonic QPU offers
Developed by Quandela, a French quantum computing leader founded in 2017. Their technology is based on Linear Optical Quantum Computing (LOQC), utilizing proprietary single-photon sources and reconfigurable optical circuits to process information at room temperature. Available through Perceval and Merlin SDK.
| Photonic QPU name | QPU model | Qubits & topology | Fidelity metrics | Speed metrics | Price per mode |
|---|---|---|---|---|---|
| QPU-ASCELLA-6PQ | Quandela Moisaiq-6 | 6 photons, 12 modes All-to-All | 1 gate: 99.6% 2 gates: 99% readout: 99% | shots: 4Mhz run: 144 op/s | 0.3€/circuit +0.000001€/shot or 750€/hour |
| QPU-ALTAIR-10PQ | Quandela Moisaiq-10 | 10 photons, 20 modes All-to-All | 1 gate: 99.94% 2 gates: 98,2% readout: 99% | shots: 3Mhz run: 400 op/s | 0.3€/circuit +0.000001€/shot |
| QPU-BELENOS-12PQ | Quandela Moisaiq-12 | 12 photons, 24 modes Dual Rail Encoding All-to-All | 1 gate: 99.6% 2 gates: 99% readout: 99% | shots: 3Mhz run: 576 op/s | 0.3€/circuit +0.000001€/shot or 1000€/hour |
Photonic emulation offers
Developed by Quandela, exQalibur is a cutting-edge photonic quantum emulator accelerated by Scaleway's most powerful GPUs. This synergy enables large-scale simulations, allowing users to explore complex parameter spaces across 31 photonic qubits at kilohertz rates, accelerating prototyping and optimization of advanced quantum algorithms. Available through Perceval and Merlin SDK.
| Photonic emulation name | Hardware & emulator | Qubits & topology | Fidelity metrics | Price |
|---|---|---|---|---|
| EMU-SAMPLING-L4 | L4 GPU exQalibur | 26 photons All-to-All | Logical qubit | 1.125€/hour |
| EMU-SAMPLING-2L4 | 2x L4 GPU exQalibur | 27 photons All-to-All | Logical qubit | 2.25€/hour |
| EMU-SAMPLING-4L4 | 4x L4 GPU exQalibur | 28 photons All-to-All | Logical qubit | 5.5€/hour |
| EMU-SAMPLING-4H100SXM | 4x H100 SXM GPU exQalibur | 30 photons All-to-All | Logical qubit | 17.415€/hour |
| EMU-SAMPLING-8H100SXM | 8x H100 SXM GPU exQalibur | 31 photons All-to-All | Logical qubit | 34.542€/hour |
| EMU-ASCELLA-6PQ | L4 GPU exQalibur | 6 photons, 12 modes All-to-All | 1 gate: 99.6% 2 gates: 99% readout: 99% | Free |
| EMU-ALTAIR-10PQ | L4 GPU exQalibur | 10 photons, 20 modes All-to-All | 1 gate: 99.94% 2 gates: 98,2% readout: 99% | Free |
| EMU-BELENOS-12PQ | L4 GPU exQalibur | 12 photons, 24 modes Dual Rail Encoding All-to-All | 1 gate: 99.6% 2 gates: 99% readout: 99% | Free |
Neutral-atoms offers
Pasqal specializes in neutral atom quantum computing, controlling Rydberg atoms with high-precision optical tweezers. Their technology allows for arbitrary 2D and 3D topology and operates in Analog mode, which is particularly well suited for large-scale quantum simulation and optimization problems involving up to a hundred qubits. Available through Pulser SDK.
| Neutral-atoms QPU name | QPU model | Qubits & topology | Speed metrics | Price |
|---|---|---|---|---|
| QPU-FRESNEL-100PQ | Pasqal Orion Beta | 100 atoms 2D layout | shots: 0.25hz | 3.3€/shot (0.83€/sec) |
| QPU-DISTRIQ-100PQ | Pasqal Orion Beta | 100 atoms 2D layout | shots: 0.25hz | 3.3€/shot (0.83€/sec) |
Neutral-atoms emulation offers
We provide access to high-performance emulation environment, leveraging DGX-class GPU clusters to create digital twins of Pasqal QPUs. This is a crucial tool for reliably prototyping and validating complex Analog pulse sequences before running them efficiently on the physical hardware. Available through Pulser SDK.
| Neutral-atoms emulation name | Hardware & emulator | Qubits & topology | Speed metrics | Price |
|---|---|---|---|---|
| EMU-FRESNEL-100PQ | A100 DGX cluster Pulser emulator | 100 atoms (tensor network) 2D layout | shots: 0.25hz | 11.2€/hour |
Superconduting QPU offers
IQM, a European leader founded in 2018, specializes in Superconducting quantum computers based on transmon technology. These processors are optimized for the general-purpose, universal gate model, making them ideal for the design and execution of standard algorithms, providing high reliability in a low-temperature environment. Available through Qiskit, Pennylane and CUDA-Q™ SDK.
| Superconducting QPU name | QPU model | Qubits & topology | Fidelity metrics | Speed metrics | Price per mode |
|---|---|---|---|---|---|
| QPU-SIRIUS-24PQ | IQM Quantum Computers Star-24 | 16 qubits (8/24 broken qubits) Central resonator | 1 gate: 99.89% 2 gates: 98.27% readout: 98.05% | TBD | 0.2€/circuit +0.00075€/shot or 1200€/hour |
| QPU-GARNET-20PQ | IQM Quantum Computers Crystal-20 | 20 qubits Square lattice | 1 gate: 99.88% 2 gates: 98.40% readout: 96.80% | run: 2600 op/s | 0.22€/circuit +0.0012€/shot or 2000€/hour |
| QPU-EMERALD-54PQ | IQM Quantum Computers Crystal-54 | 54 qubits Square lattice | 1 gate: 99.80% 2 gates: 98.86% readout: 96.53% | run: 2550 op/s | 0.25€/circuit +0.0014€/shot or 3000€/hour |
Trapped-ions QPU offers
The IBEX-Q1 by Alpine Quantum Technologies (AQT) brings trapped-ion technology to the QaaS portfolio. Known for its extremely high fidelity and long coherence time, the IBEX QPU offers all-to-all connectivity between its qubits, simplifying complex circuit design. The device is rack-mountable, operates at room temperature, and consumes less than 2 kilowatts, ideal for precision-critical simulations (like fluid dynamics). Available through Qiskit, Cirq, Pennylane and CUDA-Q™ SDK.
| Trapped-ions QPU name | QPU model | Qubits & topology | Fidelity metrics | Speed metrics | Price per mode |
|---|---|---|---|---|---|
| QPU-IBEX-12PQ | AQT IBEX-Q1 | 12 trapped-ions All-to-All | 1 gate: 99.97% 2 gates: 98.7% | shots: 5.5hz | 0.3€/circuit +0.021€/shot |
Trapped-ions emulation offers
By leveraging the power of Qiskit Aer and dedicated GPU resources, we provide digital twins of the IBEX-Q1 quantum computer. This emulation accurately models the all-to-all connectivity and topology of the trapped-ion QPU, offering accessible, cost-effective environments for prototyping and experiment. Available through Qiskit, Cirq, Pennylane and CUDA-Q™ SDK.
| Trapped-ions emulation name | Hardware & emulator | Qubits & topology | Fidelity metrics | Price |
|---|---|---|---|---|
| EMU-IBEX-12PQ-16C-128M | 16 vCPU & 128GB Aer 0.17 | 12 trapped-ions All-to-All | 1 gate: 99.97% 2 gates: 98.7% | 0.82€/hour |
| EMU-IBEX-12PQ-L4 | L4 GPU Aer 0.17 | 12 trapped-ions All-to-All | 1 gate: 99.97% 2 gates: 98.7% | 0.75€/hour |
General purpose emulation offers
Scaleway offers diverse, managed and hardware-agnostic emulation engines optimized for HPC.
NVIDIA CUDA-Q™ runtime is a fast GPU-optimized emulation engine that deliver state-of-the-art performances up to 38 qubits on B300 GPUs. Accessible through Qiskit, Cirq, Pennylane and CUDA-Q™ SDK.
| Emulation name | Hardware & emulator | Qubits & topology | Fidelity metrics | Price |
|---|---|---|---|---|
| EMU-CUDAQ-64C-512M | 64 vCPU & 512GB CUDA-Q™ 0.13 | 34 qubits (statevector) All-to-All | Logical qubits | 3.3€/hour |
| EMU-CUDAQ-H100 | H100 GPU CUDA-Q™ 0.13 | 33 qubits (statevector) All-to-All | Logical qubits | 2.52€/hour |
| EMU-CUDAQ-2H100SXM | 2x H100 SXM GPU CUDA-Q™ 0.13 | 34 qubits (statevector) All-to-All | Logical qubits | 6.018€/hour |
| EMU-CUDAQ-4H100SXM | 4x H100 SXM GPU CUDA-Q™ 0.13 | 35 qubits (statevector) All-to-All | Logical qubits | 11.61€/hour |
| EMU-CUDAQ-8H100SXM | 8x H100 SXM GPU CUDA-Q™ 0.13 | 36 qubits (statevector) All-to-All | Logical qubits | 23.028€/hour |
| EMU-CUDAQ-8B300SXM | 8x B300 SXM GPU CUDA-Q™ 0.13 | 38 qubits (statevector) All-to-All | Logical qubits | 60€/hour |
Qiskit Aer provides powerful backend options such as statevector, density matrix, and tensor network methods compatible with multi GPU infrastructure up to 36 qubits. Accessible through Qiskit, Cirq, Pennylane and CUDA-Q™ SDK.
| Emulation name | Hardware & emulator | Qubits & topology | Fidelity metrics | Price |
|---|---|---|---|---|
| EMU-AER-16C-128M | 16 vCPU & 128GB Aer 0.17 | 32 qubits (statevector) All-to-All | Logical qubits Noise model support | 0.82€/hour |
| EMU-AER-32C-256M | 32 vCPU & 256GB Aer 0.17 | 33 qubits (statevector) All-to-All | Logical qubits Noise model support | 1.65€/hour |
| EMU-AER-64C-512M | 64 vCPU & 512GB Aer 0.17 | 35 qubits (statevector) All-to-All | Logical qubits Noise model support | 3.3€/hour |
| EMU-AER-2L40S | 2x L40S GPUs Aer 0.17 | 34 qubits (statevector) All-to-All | Logical qubits Noise model support | 2.8€/hour |
| EMU-AER-4L40S | 4x L40S GPUs Aer 0.17 | 35 qubits (statevector) All-to-All | Logical qubits Noise model support | 5.6€/hour |
| EMU-AER-8L40S | 8x L40S GPUs Aer 0.17 | 36 qubits (statevector) All-to-All | Logical qubits Noise model support | 11.20€/hour |
For extreme speed, QuantumLib Qsim is a full wave function simulator written in C++ that uses advanced techniques like gate fusion, AVX/FMA vectorized instructions, and multi-threading (OpenMP) to deliver state-of-the-art quantum circuit execution up to 36 qubits. Accessible through Qiskit, Cirq and Pennylane.
| Emulation name | Hardware & emulator | Qubits & topology | Fidelity metrics | Price |
|---|---|---|---|---|
| EMU-QSIM-8C-64M | 8 vCPU & 16GB Qsim 0.21 | 32 qubits (statevector) All-to-All | Logical qubits | 0.41€/hour |
| EMU-QSIM-16C-128M | 16 vCPU & 128GB Qsim 0.21 | 33 qubits (statevector) All-to-All | Logical qubits | 0.82€/hour |
| EMU-QSIM-32C-256M | 32 vCPU & 256GB Qsim 0.21 | 34 qubits (statevector) All-to-All | Logical qubits | 1.65€/hour |
| EMU-QSIM-64C-512M | 64 vCPU & 512GB Qsim 0.21 | 35 qubits (statevector) All-to-All | Logical qubits | 3.3€/hour |
| EMU-QSIM-L40S | L40S GPU Qsim 0.21 | 32 qubits (statevector) All-to-All | Logical qubits | 1.4€/hour |
| EMU-QSIM-H100 | H100 GPU Qsim 0.21 | 33 qubits (statevector) All-to-All | Logical qubits | 2.52€/hour |
| EMU-QSIM-4H100SXM | 4x H100 SXM GPU Qsim 0.21 | 34 qubits (statevector) All-to-All | Logical qubits | 11.61€/hour |
| EMU-QSIM-8H100SXM | 8x H100 SXM GPU Qsim 0.21 | 36 qubits (statevector) All-to-All | Logical qubits | 23.028€/hour |
No. While understanding quantum principles helps, our platform is designed for developers. We support standard, high-level open-source SDKs (like Qiskit, Cirq, and PennyLane) that abstract much of the physical complexity. You can start coding in Python immediately using familiar libraries.
Real quantum hardware is a scarce and valuable resource. Our GPU-accelerated emulators act as a "Digital Twin," allowing you to prototype, debug your code, optimize circuit depth, and validate results at no noise. This ensures that when you do reserve a real QPU, every shot counts.
Logical qubit is an abstract information storage built to be fault tolerant and time resilient to quantum operations. It is built on a set of hundreds or thousands of physical qubits. Physical qubits tend to be closer to the quantum material and more unstable.
We offer a flexible model tailored to your usage.
Yes. Scaleway is a European cloud provider. Your data, your code, and your results are hosted in our data centers within the European Union, protected by strict EU privacy laws. We provide a sovereign alternative to non-EU hyperscalers, essential for sensitive research and industrial applications.
Absolutely. This is the core of our "hybrid" approach. You can trigger quantum jobs from your existing Scaleway Instances or Serverless Functions, store input/output data in Object Storage, and process the results using high-performance classical CPUs or GPUs.
For emulation and "pay-per-shot" QPU access, provisioning is almost instant. For dedicated QPU reservations (Quandela or IQM), slots are booked via our console. We recommend booking in advance for specific hardware to guarantee availability for your experiments.
