Qiskit Runtime: A Cloud-Native, Pay-As-You-Go Service for Quantum Computing

6 min read

Today, organizations are faced with the challenge of solving larger and increasingly complex problems faster, cheaper or more accurately to meet the growing needs of their industry.

This has driven them to increasingly rely more on computational methods and, therefore, require compute services as a part of solving problems. Over the last decade, this need has expanded beyond the elasticity of compute to requiring increasingly more powerful and specialized compute to solve such problems. This need has led to the availability of graphical processors (GPUs) and, more recently, experimental quantum systems (QPUs) on the cloud.

IBM Quantum is developing powerful quantum systems and services needed to enable organizations to tackle the increasing complex problems facing their industries more efficiently. Getting there requires higher-performance devices, more sophisticated development tools and — crucially — expanding the pool of developers who can use these processors, some of whom might not have a physics degree. Building this future requires the maturation of cloud-based quantum computing services.

Since IBM put the first quantum computer on the cloud in 2016, a community of 400,000 users has started exploring the frontiers of this nascent technology. These users employ the open-source Qiskit software development kit as the tool for developing and compiling quantum circuits. As quantum computer users' needs evolve, however, running single circuits through a circuit API can limit productivity with long latencies and inefficiency in abstraction. Thus, we've expanded Qiskit with the Qiskit Runtime API and programming model to handle more complex workloads. 

A majority of our quantum users today are kernel developers — those focused on studying the quantum system who require lower-level primitive programs to customize their workflows and notably, who are familiar with the intricacies of noise and mitigating errors. However, as our community grows, we have seen an increasing emergence of algorithm developers who want to focus on solving domain problems by building quantum programs that use classical computers and quantum computers together to optimize their quantum workload. These developers are more concerned with how to interpret and use the result outputs of their circuits as they leverage the tradeoff between speed and accuracy to tune existing error mitigation and suppression techniques developed by kernel developers. 

Qiskit Runtime is meant to expand the Qiskit programming model with a set of primitive interfaces that deliver modular abstraction to better support the varying needs of these two developer types. As Qiskit Runtime evolves, we expect an influx of new users looking for easy access to the highest performing quantum computing systems, with requirements ranging from complete automation to abstracted levels of noise management on quantum systems. We must support both our current community and these incoming users. 

We hope to provide a frictionless quantum computing experience where researchers and developers across both previously mentioned audiences can reap the benefits of quantum computing intuitively.

Quantum computing and the path to the cloud

If you count yourself among these new users, you may need a quick introduction to quantum computing. Quantum computers are processors built on a completely different architecture from classical computer processors, employing the principles of quantum mechanics to run a subset of algorithms that would be inaccessible even to today's most powerful supercomputers. We expect that in the coming years, quantum computers could provide substantial speed increases for problems in chemistry, machine learning, energy and more. 

IBM Quantum is a leader in cloud-based quantum computing. The team put the first quantum computer on the cloud in 2016 — a landmark that made quantum computers available to a broad audience for the first time. The next year, we introduced an open-source, Python-based software development kit to program these processors called Qiskit

Today, IBM Quantum is pushing forward on building the highest-performing quantum processors as measured by their scale, quality and speed. We measure scale by the number of quantum bits, with the 127-qubit Eagle released last year. We measure quality with the Quantum Volume metric, which calculates the accuracy of these processors' outputs. We measure speed with the recently introduced Circuit Layer Operations Per Second metric, or CLOPS. All the while, Qiskit has matured into a global community of researchers, developers and learners programming quantum computers at both the circuit level and the application-module level. 

Guiding the development of quantum computers is our technology roadmap. This document sets ambitious timelines for our hardware and software to realize Quantum Advantage — that is, applications for which quantum computation can outperform classical computation, and frictionless quantum computing or quantum computing that doesn't require an intimate knowledge of hardware and quantum noise in order to run applications.

Welcoming new developers with Qiskit Runtime

We now sit at an inflection point in the history of quantum computing, as hardware and software have matured enough to welcome a new audience of developers to the field. Quantum programs typically incorporate both quantum and classical elements, often with many passes back and forth between the two. However, previous compute flows led to inefficiencies (such as accumulated latencies), and scaling was inefficient, as running applications required the user to understand the behavior of the processor on which they ran their code. But the team has completely re-envisioned this workflow with the Qiskit Runtime API.

The Qiskit Runtime service is built with a containerized execution environment and programming model that allows our users to optimize workloads and efficiently execute them on cloud-based computing systems at scale. Qiskit Runtime lets users deploy programs instead of circuits. Containers allow them to package up code and all its dependencies so that the quantum program runs quickly and reliably from one computing environment to another.

This model is designed to not only optimize iterative workloads that require rapid parameter and setting updates, but also enable the rapid provisioning and deprovisioning of classical resources to host near-time pre- and post-processing of quantum programs to deliver error mitigation and suppression as managed services. With this iteration, Qiskit Runtime set the stage for slashing latencies and programming without deep knowledge of quantum hardware.

Qiskit Runtime is now available as a quantum computing service on IBM Cloud, allowing users to run programs on larger quantum computers (that were previously limited to specific clients) as a pay-as-you-go model and to access classical developer tools and complementary cloud services. This not only makes quantum systems more accessible to the broader user base, but it also sets the stage for users to weigh cost in the building and distributing of workloads across classical and quantum components of a workflow. And, to make quantum computing even easier for algorithm developers, Qiskit Runtime now provides primitives, a collection of core functions that enable a larger subset of our userbase to work with these distributions easily and efficiently. 

The first two primitives available are Estimator and Sampler. Estimator allows users to efficiently calculate expectation values of operators, which can be used in various applications like representing electronic energy structure, the magnetization of a spin material or the kernel of a machine learning problem. Sampler takes a quantum circuit and outputs a probability distribution. This is useful for search applications like Grover's algorithm.

We've already begun reaping the benefits of Qiskit Runtime; last year, we reported seeing speed-ups of up to 120x on some workloads, with Qiskit Runtime providing a sizable contribution to those results.

The road to frictionless quantum computing

Qiskit Runtime on IBM Cloud does more than expand the developer audience. Our vision of frictionless quantum computing is one where users work in an obstacle-free cloud computing landscape, leveraging Qiskit Runtime seamlessly alongside other cloud compute and storage services. To experiment with real problems at scale, users need to deploy and manage elastic, heterogeneous workloads efficiently. We hope that, soon, users will be able to consume quantum in a serverless fashion with elastic and specialized classical resources, with no need for infrastructure management consideration, where scaling will be a seamless process so users can focus simply on working toward Quantum Advantage. Users need only write code and upload it.

Qiskit Runtime begins to open a door toward this future. Any IBM Cloud developer can use Qiskit Runtime to access larger hardware — previously restricted to specific clients — with their existing IBM Cloud credits or a credit card, paying only for the quantum services they need. Users can begin to imagine building tools that incorporate quantum computing as any other program would incorporate GPU resources. This also includes organizations building services on top of the Qiskit Runtime service. For example, Strangeworks' developer platform integrates Qiskit Runtime Services as an embedded solution through IBM Cloud.

We plan to continue toward our vision for frictionless quantum computing by working tirelessly along our development roadmap. We will continue to use agile processes to push the scale, quality and speed of our processors, and in the next two years, we plan to release our 433-qubit Osprey and 1,121-qubit Condor processors.

We're continuing to explore applications, using Qiskit Runtime to introduce more primitives and make quantum computing more accessible for an even wider class of developers we call model developers. We're coupling quantum and classical computing more closely, including using real-time classical resources to run dynamic circuits, and developing techniques that employ high-performance computing to split up larger problems to more intelligently leverage classical and quantum resources — potentially accelerating the path toward Quantum Advantage.

Get started with Qiskit Runtime

Our vision is to set the groundwork for quantum computing to solve some of society's hardest problems. Qiskit Runtime, combined with the continued progress along our roadmap, sets the stage for a revolutionary rethinking of quantum computation, employing the best practices of cloud computing to bring quantum computing to the broadest community of developers. With Qiskit Runtime serving as the nexus of the quantum computing experience, we hope to see industries around the globe begin reaping the benefits of quantum in the coming years.

Ready to see how Qiskit Runtime can optimize your workloads? Get started today.

Read more about Qiskit Runtime: "Qiskit Runtime primitives make algorithm development easier than ever"

Check out the "Qiskit Runtime explainer."

Be the first to hear about news, product updates, and innovation from IBM Cloud