D-Wave “Supremacy” Controversy Overshadows Real Progress

Quantum-computing company D-Wave sparked controversy earlier this month by saying it could solve problems beyond classical computers’ capabilities. The claim was quickly challenged, but despite the furor, experts say the results demonstrate the growing capabilities of its technology.

“Quantum supremacy,“ which refers to the point when a quantum computer can solve a problem that is essentially impossible for a classical computer, has become the guiding star for the quantum-computing industry. Google, startup Xanadu, and researchers at the University of Science and Technology of China have all claimed to have crossed this milestone. But they have been criticized for using tests that are heavily biased toward the quantum hardware, as they essentially get these processors to produce a random output and then challenge classical computers to simulate it.

D-Wave, headquartered in Palo Alto, Calif., is now the latest company to claim quantum supremacy, by way of results published 12 March in the journal Science. But this time the experiments involve quantum simulations that the company says have potential value for both scientific research and materials discovery. D-Wave’s claims have faced pushback from experts who specialize in simulating quantum systems on classical hardware, but the company is sticking to its guns.

“Our claim is this machine can simulate some problems that cannot be solved classically and answer some scientific questions that can never be answered classically,” says Mohammad Amin, chief scientist at D-Wave. “This claim, I think, will always survive.”

While companies like Google, IBM, and IonQ are building “universal quantum computers” that can, in principle, run any quantum algorithm, D-Wave’s machines rely on a process called quantum annealing. This approach can tackle only certain types of tasks, most prominently optimization problems. It involves carefully programming the parameters of a problem into the machine’s qubits and then letting them evolve over time to find a solution.

In its latest experiments, D-Wave’s researchers used a prototype of its new Advantage2 processor featuring roughly 1,200 superconducting qubits to simulate the quantum dynamics of magnetic materials with different structures and sizes over various time scales. They then worked with external researchers who attempted to replicate the results using leading classical techniques. In the Science paper, the authors claim that matching D-Wave’s results on the largest problems would take the best classical approach, known as matrix-product states (MPS), millions of years on the Frontier supercomputer at Oak Ridge National Laboratory.

D-Wave vs. Classical Computing Controversy

D-Wave’s assertion of quantum supremacy was quickly challenged. Researchers from the École Polytechnique Fédérale de Lausanne, in Switzerland, demonstrated that they could carry out calculations that D-Wave’s estimates suggest would take more than 150 years on Frontier in just a few days using four graphics processing units. Another group from the Flatiron Institute in New York City showed that combining MPS with another algorithmic technique allowed them to simulate some of the smaller models in just hours on a single CPU. They also showed that the approach could be readily scaled to larger problems.

Both studies push the state of the art for classical techniques, says Juan Carrasquilla, an associate professor of physics at ETH Zurich, in Switzerland, and a coauthor on the Science paper, who carried out some of the calculations that D-Wave benchmarked against. But neither study overturns the paper’s central claim because both groups tackled only a subset of the problems the company simulated, and at shorter time scales, he says. And while on paper these approaches look scalable, in practice it can be more challenging than it seems.

“This narrows this range of the claim, but if you want to completely overturn it you should be able to simulate the systems that D-Wave presented in the paper,” says Carrasquilla.

Miles Stoudenmire, a research scientist at the Flatiron Institute, says his group didn’t see scientific value in extending their approach to bigger problems, but that they now plan to do so to prove the point. More broadly though, he takes issue with D-Wave’s claim of supremacy because it ignores the possibility of further advances in classical approaches.

“The core problem is that their claim not only isn’t true, but it can’t be true in some sense because it’s a claim about the future,” he says. “We find that very off-putting and, frankly, unscientific language.”

Supreme or Not, the Results May Be Useful

The controversy around whether D-Wave’s machine can outstrip the fastest imaginable computer also overshadows a more pertinent fact—it is much faster than today’s best classical techniques. While there are other trade-offs to consider, such as the cost and the inflexibility of the system, Stoudenmire says problems that take hours to solve using his approach can be tackled in fractions of a second by D-Wave’s machine.

Filippo Vicentini, a professor of AI and condensed-matter physics at the École Polytechnique, in Paris, says that even if it can’t demonstrate supremacy, D-Wave’s machine could still have practical advantages. “You don’t want to run your GPUs for hundreds or thousands of hours if you have a quantum processor that does it in a few minutes,” he says.

However, while there may be scientific questions that can be answered by simulating the kind of quantum dynamics featured in the Science paper, there are no obvious commercial applications, says Vicentini. In a 2023 Nature paper, D-Wave showed that some optimization problems could be mapped onto this type of quantum simulation, and that its machine could do these simulations quicker than classical computers. But Vicentini points out that there are a large number of powerful classical optimization algorithms that can solve these problems directly, without first converting them to quantum simulations.

Carrasquilla agrees, and also notes that media coverage suggesting the results could have practical applications is misleading. “When people say we can now simulate materials, that refers to a very, very small corner of materials, and definitely not molecules, not drug discovery, not materials discovery,” he says. “It’s very interesting for statistical mechanics and quantum many-body physics, but for commercially relevant problems I think this is still far away.”

Carleton Coffrin, a staff scientist at Los Alamos National Laboratory, is more optimistic. He leads a project to use D-Wave’s computer to simulate “quantum magnets,” which could one day form the basis of ultraprecise sensors or memory devices for quantum computers. The ability to run experiments that take days on a supercomputer in just minutes is proving particularly powerful. “That allows the condensed-matter theory expert to interact with the hardware in almost real time,” Coffrin says. “That just dramatically accelerates the type of insights you get.”

The group has also been testing D-Wave’s quantum computers on optimization problems since 2017. The quantum simulations reported in Science have no direct application to this class of problems, says Coffrin, but his group has nonetheless seen significant jumps in performance each time D-Wave has upgraded its underlying qubit technology. In 2022, they reported that on certain optimization problems, an older generation D-Wave computer featuring roughly 5,500 qubits was 15 times as fast as the best classical algorithm.

That kind of gap could probably be closed by a concerted effort to develop new classical algorithms. But if the company can double their qubit count to 10,000, Coffrin suspects it will become increasingly difficult for classical approaches to keep up. “At that scale, all other things being equal, the classical competition would have a long way to go,” he says. “It would be hard.”

This article was updated on 3 April 2025 to clarify the work done by Juan Carrasquilla.