by Mena Issler, Henning Soller, and Matija Zesko
Quantum computing promises eye-popping progress, and it’s expected to have an economic impact of almost $1.3 trillion by 2035. However, that value is not yet within reach: fully fault-tolerant quantum computers that can generate significant business value are still years down the road.
While we wait, quantum optimizers may be a bridge technology.1 While quantum optimizers won’t be able to solve as many problems as quantum computers are expected to solve, they are easier to build and computationally superior to classical computers, at least for some optimization problems.
In May 2022, a functional quantum optimizer demonstrated its advantage against high-performance classical computers for a set of problems that are inherently difficult for classical computers.2 Experts believe this quantum optimizer can deliver commercial value in solving optimization problems across industries, in forecasting, and in materials science.3
The full potential of this quantum optimizer will become clear as more users in academia and industry experiment and apply it to different use cases. In this blog post, we shed light on possible applications and discuss what decision makers can consider in evaluating the potential use of these applications for their organizations. The technology could offer a jump-start on harvesting the rewards of quantum computing.
Technical advantages and considerations
On the technical side, quantum optimizers can solve problems that are difficult for classical computers. Using a technology based on neutral atoms, quantum optimizers arrange atoms in a grid and excite them to simulate a computational problem. This process can produce an approximate—but accurate enough—result in a very short time.
Commercially, quantum optimizers are more scalable and cost-effective than universal quantum computers.
These advantages mean they are likely to remain competitive against both classical and quantum computers, even after universal quantum computers become viable. Of course, universal quantum computers would be able to solve a wider array of problems than optimizers, but quantum optimizers could scale much more quickly, because unlike universal quantum computers, their functionality does not depend on individual control of each of the more than one million quantum bits (qubits). Indeed, universal quantum computers will require complex designs that have not yet been developed and that quantum optimizers don’t need.
The kinds of use cases that can be implemented for quantum optimizers are currently restricted by the number of available qubits in devices (a maximum of about 250 as of early 2023). As early as 2021, companies were promising to release a 1,000-qubit quantum optimizer by 2023,4 and experts we interviewed expect qubits to scale up rapidly in the following years. This will allow quantum optimizers to solve problems with an increasing number of variables.
The recently developed neutral-atom quantum optimizer’s advantage with optimization problems makes it useful in industries such as telecommunications, logistics, autonomous driving, and pharmaceutical research.
In the short term, some of the most applicable optimization problems that quantum optimizers could help solve are those that can be mapped onto graphs (mathematical structures that represent relationships between pairs of objects). Atoms could represent the nodes of the graph, and the distances between atoms on the grid could represent the edges. Researchers are continually developing new ways of mapping problems onto quantum optimizers,5 so new classes of optimization problems could be unlocked soon.
One caveat: the time required to find an approximate solution to an optimization problem increases with more variables. Therefore, the most significant near-term benefit from quantum optimizers will likely be found in problems in which the quality of the output matters more than the speed of getting the result. Relevant use cases are problems in which a slight improvement can create significant cost savings, such as antenna placement in telecommunications and footprint optimization in retail. In these cases, atoms’ positions in a grid could simulate the optimal geographical placement.
As the number of qubits grows over time, quantum optimizers are expected to become faster when compared with classical optimizers, and to solve in seconds problems that would take classical devices minutes or hours to solve. This makes the technology useful for problems that are time sensitive or that require frequent recalculation because of dynamic variables—for example, making real-time calculations to find the optimal route for autonomous vehicles or to enable 5G mobile networks in crowded areas. Telecommunications, particularly the Internet of Things, involves many similar problems that focus on sending information efficiently through dynamically changing networks.
Possible next steps
Decision makers can investigate ways to use quantum optimizers to generate value now before fault-tolerant quantum computers come online. The first step is to identify optimization problems in their businesses that cannot be easily solved with classical computers. From there, organizations could scope the extent of the challenge, estimate the potential value to be gained, and experiment with quantum optimizers to confirm a fit between the problem and a possible solution. After that, cracking previously unsolvable problems will be a matter of collaborating with providers of quantum optimizers to design and run the right tools.
Quantum optimizers can be a valuable waypoint toward universal quantum computers. Early value from quantum computing might actually be attainable now.
Mena Issler is an associate partner in McKinsey’s San Francisco office, Henning Soller is a partner in the Frankfurt office, and Matija Zesko is a consultant in the Zurich office.
1 For a primer on quantum optimizers, see Peter Byrne, “Analog simulators could be shortcut to universal quantum computers,” Scientific American, May 6, 2015.
2 For more on the optimizer, see B. Barak et al., “Quantum optimization of maximum independent set using Rydberg atom arrays,” Science, May 2022, Volume 376, Number 6598.
3 For more on commercial applications, see Nathan Gemelke et al., “Industry applications of neutral-atom quantum computing solving independent set problems,” Cornell University, May 17, 2022.
4 Maija Palmer, “Pasqal raises €25m to build 1,000-qubit quantum processor,” Sifted, June 8, 2021.
5 For example, see Lin-Guo Liu et al., “Quantum optimization with arbitrary connectivity using Rydberg atom arrays,” Cornell University, March 2, 2023.