OVH and Atos face the American giants


While waiting for the first commercial quantum computer, cloud providers are offering to run quantum algorithms on their platform. An “as a service” approach that allows you to experience the announced revolution.

Quantum computing remains a promise that will only be realized in about ten years. While waiting for the actual release of the first commercial quantum computers, companies can now interact with this disruptive technology. IBM and the three American hyperscalers offer Quantum Computing as a Service (QCaaS) offerings. In France, this is also the case for OVHCloud in partnership with Atos. Based on the simulation or emulation of quantum computing infrastructure, these solutions allow developers to become familiar with writing future applications today. “These cloud services are mainly used to learn how to code and test algorithms and not to put projects into production”, specified Olivier Ezratty, consultant specializing in quantum technologies and author of the ebook which is Understanding Quantum Technologies, which is updated annually. . “So companies are putting their feet in quantum and confusing it. On the supplier side, it’s a question of occupying the land and setting a date.”

Comparison of quantum clouds
IBM Quantum Amazon Bracket Azure Quantum Google QuantumAI Atos and OVH Cloud
Year of launch 2016 2019 2019 2018 2022
Quantum computers More than 20 own quantum machines IonQ, Rigetti, D-Wave… IonQ, QCI, Rigetti, Quantinum… own quantum computer own quantum computer
Power (number of Qubits) From 5 to 127 From 12 to 50 CN Up to 54 Until 38
Runtime environment Squirt Amazon Bracket SDKs, Jupyter Notebook, Pennylane Quantum Development Kit (QDK), Cirq, Qiskit, Q#, Jupyter Notebook Cirq, OpenFermion, TensorFlow Quantum Jupyter Notebook
Sources ExxonMobil, JPMorgan Chase, Samsung, Boeing, Sony, Capgemini… Fidelity, the Italian National Institute of Nuclear Physics (INFN), Pasqal… NASA, OTI Luminonics, Willis Towers Watson… CN CN
Prices Flat-rate or pay-per-use. Starting at $1.60 per second of execution. From $0.30 per job and $0.00019 per implementation cost (D-Wave Advantage) $500 free credit per supplier. Monthly subscription or pay-as-you-go. From $0.00003, 1-qubit gate execution (IonQ Harmony) CN CN

IBM opened up the market by offering, in 2016, a quantum computer service in the cloud. Big Blue relies on its own machines that display power of 15 to 127 qubits. As a reminder, a qubit is the unit of measurement of information in quantum computing. According to the principle of superposition of states, specifically in quantum physics, this quantum bit can simultaneously represent a 0 and a 1. Thus, the calculation speed increases exponentially, the machine is able to simultaneously process of several states at that time. Measuring qubit power? This is 2 to the power N (N is the number of qubits in the processor). Thus, a binary machine based on a 6-bit architecture can create one of 64 (2 to the power of 6) possible combinations (000000, 000001, 000010…). The qubit, which can be a superposition of 1 and 0, can suddenly go into 64 states.

“Qubits are not uniform. You have to look at their quality or the error rate”

Like IBM, Google chose to design its quantum machines. In contrast, AWS and Microsoft use third-party manufacturers and mix a wide range of solutions. “The technologies are still immature and the machines have very variable performance,” tempers Olivier Ezratty. “Qubits are not all equal. You have to look at their quality or the error rate.” To learn how to program, these vendors offer visual interfaces for building and testing circuits. They also provide developers with Jupyter notebooks, software libraries and scripting tools in Python, the common language for AI and quantum.

From our previous comparison, a French and European player, OVHCloud, entered the sector supported by Atos. The provider can be followed by other national representatives such as the very promising start-ups Alice & Bob or Pasqal, co-founded by the recent Nobel Prize in Physics Alain Aspect. Things are also moving on the public research side with the launch, in early 2022, of a national quantum computing platform, bringing together INRIA, CEA and Genci (National Intensive Computing Large Equipment).

IBM Quantum, the pioneer bonus

First to offer quantum “as a service”, IBM dominates the market by offering, from its laboratory in New York, more than 20 quantum machines. Called IBM Quantum System One, its offering is based on Falcon 27 Qubits, Hummingbird 65 Qubits and Eagle 127 Qubits processors. Big Blue doesn’t intend to stop there. Announced in May, its roadmap plans to introduce a 433-qubit processor, called Osprey, this year before launching, in 2023, Condor, “the world’s first universal quantum processor with more than 1,000 qubits”.

Also in 2023, IBM will offer a serverless approach to make the quantum development experience more accessible. In the meantime, developers can use the Qiskit execution engine to create quantum programs and run them on the prototype quantum infrastructure that IBM has made available in the cloud. Membership in the IBM Quantum Network provides support from quantum computing experts. The digital giant counts CERN, Mitsubishi Chemical or ExxonMobil among its references.

AWS, the widest range of technologies

Unlike IBM, Amazon Web Services does not have its own quantum machines. With its Amazon Bracket offering, the provider uses a variety of manufacturers. This allows to cover different technologies including quantum annealing with D-Wave, trapped ions with IonQ or superconducting quantum processors from Oxford Quantum Circuits and Rigetti. Xanadu’s quantum chip (QPU) provides access to photonic quantum computers.

As an overlay, the developer can use designs and execution environments for quantum algorithms that are agnostic to the underlying technology. Starting with the Amazon Bracket SDK. The AWS service also supports Jupyter notebooks and the PennyLane open source software framework. The Amazon Quantum Solutions Lab program connects experts in quantum computing. Pricing has two components: a cost per execution and a charge per task.

Microsoft Azure, the most open environment

Like AWS, Microsoft hasn’t built quantum machines. The cloud provider uses solutions designed by specialists based on different technologies such as captured ions (IonQ, Quantinuum), superconductors (QCI, Rigetti), the neutral atom (Pasqal) or algorithms -optimized (1QBit, Microsoft QIO, Toshiba SQBM+).

Called Azure Quantum, its offering is distinguished by the richness of its development environment. Along with its own development kit, the QDK (for quantum development kit), Microsoft supports the open-source Cirq and Qiskit SDKs, the Q# quantum programming language, and Jupyter notebooks. To experiment with this offer, Azure provides $500 free credit per vendor. The provider offers monthly subscriptions or pay-as-you-go.

Google, heading for first commercial quantum offering?

Google has developed its own quantum computer called Weber. Based on the latest generation Sycamore quantum processor, it boasts a maximum power of 54 qubits. It is located at the Quantum AI campus that opened in 2021 in Santa Barbara, California, where Google researchers are working to build the first commercial quantum computer.

For cloud workloads, Google Quantum AI offers to provide quantum virtual machines or QVM (for quantum virtual machine). For this, the hyperscaler provides several software libraries. Coded in Python, Cirq allows you to write quantum circuits and then execute them. OpenFermion is used to compile and analyze quantum algorithms to simulate fermionic quantum systems. Finally, TensorFlow Quantum focuses on hybrid machine learning, both quantum and traditional.

Atos and OVHCloud, first European cloud offer

Two French players have joined forces to offer the first European quantum cloud offer. On June 9, 2022, Atos and OVHcloud announced in a joint press release that they had entered into a partnership in the field of quantum computing. Goal: make the Atos quantum emulator available on the OVHcloud cloud through an “as a service” approach.

Based on the SMP server power of Atos’ BullSequana X800 supercomputer, the quantum learning machine (QLM) deployed on OVHcloud covers three quantum programming modes: the gated model, quantum annealing and the analog model. “Users will be able to simulate circuits up to 38 qubits in double precision and solve quantum annealing problems up to 5,000 qubits”, advance the two partners. QLM technology also paves the way for first generation NISQ or noisy intermediate-scale quantum processors. Especially systems that are still imperfect, include errors, but nevertheless can perform tasks that are inaccessible to conventional computers.

Leave a Reply

Your email address will not be published. Required fields are marked *