Quantum pioneer D-Wave is building a quantum computer to rival Google and IBM

D-Wave Systems has the distinction of being the first company to bring to market a working quantum computer, a would-be successor to today’s supercomputers. But there has always been an asterisk appended.

D-Wave’s machine, which debuted in 2011, was a special kind of quantum computer called a “quantum annealer.” The machine could only solve certain classes of mathematical problems. These problems—optimization and sampling, as they’re known—happened to have a lot of applications in business. But D-Wave’s annealer could not simulate the complex atomic-level physics and chemistry interactions that many scientists saw as one of the prime advantages quantum computers would have over conventional ones. Nor could it easily be used to factor large prime numbers, potentially cracking most of today’s most widely-used encryption algorithms—one of the great fears about the advent of full-fledged quantum computing.

In recent years, as a growing number of companies have brought to market primitive universal quantum computers, some of which have begun to do things that no conventional computer can do, D-Wave, which is based in Burnaby, British Columbia, near Vancouver, stuck to quantum annealing. The company has upgraded the power of its quantum annealer several times and is currently selling what is, by at least one metric, the most powerful quantum-based computer in existence: The machine has 5,000 quantum logic processing units, or qubits, which are the quantum computing equivalent of traditional computing’s bits.

But there’s always been that asterisk—until now. Today D-wave announced its intention to build a universal quantum computer and offer customers the ability to run computations on it through the cloud alongside its latest generation quantum annealer. This universal quantum computer is what scientists call a “gate model” machine because the qubits act as gates—controlling the flow of electric current in a traditional electrical circuit—and performing logical operations that are part of a computer program.

“We are absolutely committed to developing and delivering enhancements to the annealing system,” Alan Baratz, D-Wave’s chief executive officer, told Fortune. “But there are important questions that annealing cannot address in non-linear differential equations, in quantum chemistry, and physics simulations. So if we could also bring a gate model system, we would be the only company in the world with an annealing and a gate model system, and we could address the entire total addressable market for quantum computing.”

The company said it will use superconducting materials to build the qubits for its universal quantum computer. That is the same underlying approach that IBM, Google, the startup Rigetti Computing, and several other quantum computing initiatives use. Yet more companies, including electronics giant Honeywell and startup IonQ, are exploring qubits based on ions held in place by lasers. Still others are exploring qubits created by photons or using semiconductors. All are targeting a huge potential market for quantum computers that could revolutionize many business processes as well as allow for the creation of new types of chemicals and materials.

Although D-Wave was the first company to build a working quantum computer, it has struggled to gain commercial traction. Some researchers, most notably computer scientist Scott Aaronson at the University of Texas at Austin, faulted the company for over-hyping what its machines were capable of. (For a long time, Aaronson cast doubt on whether D-Wave’s annealer was harnessing any quantum effects at all in making its calculations, although he later conceded that the company’s machine was a quantum device.)

In the past few years, the company has also had trouble exciting investors: in March, it secured a $40 million grant from the Canadian government. But that came after The Globe & Mail newspaper reported that a financing round in 2020 had valued the company at just $170 million, less than half of its previous $450 million valuation. The company’s decision to add gate-model quantum computers to its lineup may be an acknowledgment that commercial momentum seems to be far greater for those machines than for the annealers that D-Wave has specialized in.

D-Wave already uses superconductors for the qubits in its annealer. Baratz said the company has expertise in constructing large arrays on these qubits on a single processor, stacking them in layers, that other companies working on superconducting gate-model quantum computers don’t have. The company said it would aim to have a 60 qubit gate-model machine available by 2023 or 2024.

D-Wave said it will follow this with a 1,000 qubit gate-model computer that will offer at least four qubits that are “error-corrected.” Calculation errors are a problem with all current quantum computers, one that has limited their practical value for many kinds of commercial and research applications. The problem occurs because scientists have only figured out how to keep qubits in a quantum state for relatively brief periods of time, usually just fractions of a second, and as these qubits fall out of a quantum state, they no longer produce accurate calculations.

D-Wave also announced plans to offer by late 2023 or early 2024 an even more powerful quantum annealer: one with more than 7,000 qubits, each of which is connected to 20 others. This annealer and the gate-model quantum computer are both part of a technology roadmap D-Wave calls “Clarity.” The future model also involves several enhancements to D-Wave’s software.

In addition, D-Wave said it was immediately releasing an upgraded software package to help customers use its existing quantum annealer in combination with standard computers to solve business problems. Known as a “hybrid solver,” because it runs parts of the algorithm on standard hardware and part of it on the quantum annealer, D-Wave’s newest version will enable customers to “solve larger and more complex problems with greater accuracy,” the company said. The new system could find a better solution to an optimization problem—such as how best to route delivery vans through city traffic—70% of the time, it said.

New research has shown that gate-model quantum computers may not provide much of a speed-up over classical computers for running optimization problems, Baratz said. These problems include a host of business challenges, such as finding the most efficient logistical route, figuring out how best to configure a factory floor, or how best to run a clinical trial for a new drug. But D-Wave has proven that its quantum annealer can definitely produce faster results for these kinds of problems than a traditional computer, he said.

D-Wave’s system has been used by a number of large companies, including Volkswagen, automotive components company DENSO, Lockheed Martin, and Japan’s NEC. It has also been used by the U.S. government’s Los Alamos National Laboratory. Several other companies will soon be able to announce they put D-Wave’s system into production too, Baratz said.

Quantum computers are machines that harness the strange properties of quantum mechanics to make their calculations. These include superposition, where a particle exists in two states simultaneously, and entanglement, where particles in a quantum state act on one another, even when physically distant. These two properties enable quantum information processing units, called qubits, to do things that the information processing units in standard computers, called bits, can’t do.

For instance, bits represent information as either a “0” or “1.” But qubits can represent both “0” and “1” at the same time. In a standard computer, the computer chips are designed so that each bit’s value is independent of the others in the processor. But in a quantum system, the qubits are designed to be entangled so that their states affect each other. These two properties allow a quantum computer, in theory, to process, far more information in parallel than a standard computer. That would enable these machines to be exponentially more powerful than any traditional computer.