Quantum Computing - Concepts

Reviewed on November 21, 2025

Overview

Quantum Computing

Quantum computing harnesses the principles of quantum mechanics to process information. Unlike classical computers that use bits (0 or 1), quantum computers use qubits. Qubits can exist in a state of superposition (representing both 0 and 1 simultaneously) and can be entangled. This allows quantum computers to solve specific classes of complex problems—such as optimization, material simulation, and cryptography—exponentially faster than classical supercomputers.

Hybrid Computing

Quantum computers are not standalone devices; they act as accelerators. Hybrid Computing is the practice of orchestrating workflows that combine the best of both worlds:

• Classical resources (CPU/GPU): For data pre-processing, post-processing, and managing the workflow.
• Quantum resources (QPU): For executing the specific quantum subroutines where they offer an advantage.

QPU (Quantum Processing Unit)

A QPU is the physical hardware chip that performs quantum computations. Unlike a CPU, a QPU is extremely sensitive to its environment and often requires cryogenic cooling or vacuum chambers.

• Nature: Physical hardware (Photonic, Neutral Atom, Superconducting, etc.).
• Access: Often requires reservation (Booking) due to scarcity.
• Use case: Final execution, proof of concept, searching for quantum advantage.

Emulator

An Emulator is a software program that mimics the behavior of a QPU. At Scaleway, these emulators run on classical GPU clusters.

• Nature: Software running on High-Performance Computing (HPC) infrastructure.
• Access: On-demand (instant availability).
• Use case: Prototyping, debugging, learning, and noise-free verification before deploying to a real QPU.

Physical qubits

Physical qubits are the actual hardware elements (photon, atoms...) that store and manipulate quantum information. They are susceptible to errors due to decoherence and operational imperfections.

Most of QPUs available today at Scaleway have a limited number of physical qubits, which constrains the size and complexity of quantum algorithms that can be executed directly on hardware.

Logical qubits

Logical qubits are an abstraction built on top of physical qubits using quantum error correction mechanisms. They provide a more stable representation of quantum information by encoding it across multiple physical qubits and error correction algorithms.

Specifications

The Scaleway QaaS workflow is built around three main objects: Platform, Session, and Job.

Platform

A Platform is the starting point. It represents a specific combination of hardware and the software stack required to run it. Think of it as a "configuration" or an "image" you select.

• Uniqueness: Each Platform ID corresponds to a unique pair of Provider + Technology (e.g., QPU-ASCELLA-6PQ for Quandela photonic QPU or EMU-FRESNEL-100PQ for Pasqal neutral-atom QPU emulator).
• Usage: You must choose a Platform ID to create a Session.

Session

A Session is a dedicated time interval during which you have access to a specific Platform to execute your code. It acts as a namespace for your work.

• Lifecycle: A session must be explicitly created (status: starting $\to$ running) and terminated (status: stopping $\to$ terminated).
• Persistence: Even after a session is terminated, its metadata and the results of the jobs attached to it remain readable (GET/LIST).
• Constraint: Once a session is terminated, it cannot be restarted. You must create a new one.

Job

A Job represents a single unit of work—typically a quantum circuit—submitted to a Session.

• Dependency: A job is always attached to a running session. You cannot submit a job to a stopped session.
• Immutability: Once created, a job cannot be deleted (for traceability), though it can be cancelled if it hasn't started processing yet.
• Statuses: A job transitions through several states: waiting (in queue), running (execution on QPU/Emulator), completed (results available), or error.

SDK

To program quantum computers, you typically use specialized Software Development Kits (SDKs). Scaleway QaaS supports multiple SDKs, each tailored to different quantum computing paradigms.

They globally allow you to:

• Build quantum circuits using Python.
• Execute them on various backends (emulators or real QPUs).

Qiskit

Qiskit is an open-source software development kit (SDK) widely used to program quantum computers. Originally developed by IBM, it has become a standard language for the quantum ecosystem.

Cirq

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.

Pulser

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.

Perceval

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

PennyLane

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

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.

Access modes and billing

Scaleway QaaS offers two primary ways to access quantum resources, depending on your needs for availability and cost control.

Emulator - Pay-As-You-Go (On-Demand)

Access method used only for Emulators. You can start a session instantly without prior reservation. The resources (GPUs or vCPUs) are allocated dynamically.

• Flexibility: No need to plan ahead. Submit jobs whenever you are ready.
• Dedicated: This is your dedicated instance to run emulated quantum Jobs, there is no shared queuing.
• Billing Structure: You are billed for the duration of the Session (e.g., per minute or hour), regardless of how many jobs or shots you run during that time.

QPU - Pay-As-You-Go (On-Demand)

This is the standard access mode for most use cases. Instead of reserving a machine for a specific time, you submit your jobs to a shared queue. You are billed strictly based on the volume of computations performed.

• Flexibility: No need to plan ahead. Submit jobs whenever you are ready.
• Shared Queue: Your jobs are processed in the order they are received (FIFO), sharing the QPU with other users.
• Billing Structure: The cost is calculated based on the number of circuits (jobs) submitted and the total number of "shots" (executions) requested.

Example

If you submit 2 quantum circuits, requesting 1000 shots for each circuit:

1. Circuit Fees: You pay the base fee for 2 circuits.
2. Shot Fees: You pay for 2000 shots (2 circuits × 1000 shots).

QPU - Booking (Reservation)

Because real physical QPUs are scarce resources, Booking allows you to reserve a specific hardware backend for a dedicated time slot in the future.

• Guaranteed Access: During your booked slot, the hardware is dedicated to you.
• No Queue: Your jobs bypass the public queue and are executed immediately.
• Billing Structure: You are billed for the duration of the reservation (e.g., per minute or hour), regardless of how many jobs or shots you run during that time.
• Use Case: Essential for intensive production runs, workshops, or when you need guaranteed throughput on partner hardware (like Pasqal or Quandela).