Murray Thom, vice president for quantum technology evangelism for D-Wave, knows that his company for more than two decades has played a key role in a somewhat unusual industry.
“One of the things often mentioned is that the quantum computing industry is in this really strange circumstance where you have one company, D-Wave, that’s been here for a long time building and commercializing quantum computing, and then you have that huge effort – like dozens or hundreds of other players – who are building very small-scale machines and starting up in the space,” Thom told The Next Platform. “One of the key things that differentiates us was that we started very early on surveying the entire field to make sure that we were making the choice that would get us to practical problems as quickly as possible.”
The company in its infancy embraced quantum annealing technology, even though IBM, Microsoft, Amazon, Google, and many others have pursued superconducting gate-model systems. The decision helped D-Wave become the first to commercialize quantum computing, he said, an advantage it continues to push. And while D-Wave in 2021 laid out plans to develop its own gate-model quantum system, its annealing technology remains the focal point of its work.
That work is on display this week in Boston at Qubits 2024, D-Wave’s annual user conference, where executives expounded on the momentum they have seen from recent announcements, like the 1,200-qubit prototype of its Advantage2 quantum processor unit (QPU) available in the company’s Leap cloud service and its new fast-anneal protocol in all QPUs in Leap for solving complex optimization problems more quickly.
At the show, D-Wave unveiled its latest quantum-classical hybrid quantum solver, a highly scalable tool that is available in Leap and aimed at increasingly difficult nonlinear optimization problems by supporting up to 2 million variables and constraints and with up to 10X the problem size capacity of its other solvers.
“At the high level, it’s doubling our problem-solving capability,” Trevor Lanting, D-Wave’s chief development offer, told The Next Platform before the conference began. “It’s making formulation of problems that are complicated with nonlinear relationships between variables – both in the objective function and the constraints – very, very easily handled in some sense natively with the solver. this increase in scale as well as increase in performance, that’s the key driver for the new nonlinear hybrid solver.”
D-Wave sold its first quantum computer – the D-Wave One – in 2011, and despite its longevity has taken arrows over the years from industry observers and larger rivals about the legitimacy of annealing being “real” quantum computing. This despite continued growth in recent years. In its fiscal year 2023, D-Wave saw a 22 percent year-over-year revenue jump, to $8.8 million, 133 total customers – including BASF, Deloitte, MasterCard, and Lockheed Martin – compared with 120 the year before, and a 41 percent increase in commercial revenue, which accounted for 70 percent of D-Wave’s overall revenue, compared to 60 percent in 2022.
In his keynote stage Monday, D-Wave president and chief executive officer Alan Baratz pushed back at the company’s critics, noting the quantum computing services and products it has available, innovations coming with Advantage2, performance advantages of its annealing technology in running optimization workloads, and the continued embrace by enterprises of the technology.
“We are seeing customers actually transitioning applications into production uses,” Baratz said. “Today – not five years from now, but today – we have the systems. We have the services today that businesses can use to leverage quantum in support of their business operations.”
Recent advances in both its annealing and still-in-development gate-model efforts are proof points of D-Wave’s lead over its rivals, which are still developing their gate-model systems, he said, outlining performance wins when running optimization workloads like workforce scheduling and supply chain applications.
“Optimization is a really important problem class,” he said. “It’s low-hanging fruit for the use of quantum today. D-Wave’s annealing quantum computers are excellent at solving this class of problems. Annealing quantum computers cannot solve all problems. For example, they’re not particularly good at quantum chemistry or designing longer lasting batteries. Gate-model systems are much better for applications like that.”
There is a bifurcation in the application environment for quantum, Baratz said. Some will always run on annealing systems rather than their gate-model counterparts, others will run best on gate models. Enterprises need to keep this in mind when considering the right quantum technology for their particular problems.
That said, D-Wave is pushing ahead with enhancements to its Advantage QPUs. The current Advantage system includes more than 5,000 qubits, but new capabilities being plugged into Advantage2 will allow it to jump ahead in a number of areas. An early Advantage2 prototype launched a couple of years ago and available in Leap has 500 qubits. It had a new architecture with qubit connectivity, but a small number of qubits and was fabricated using the same process as that used to build Advantage.
In early 2023, D-Wave created a 400-qubit version of the same architecture but it came with a new fabrication stack that included higher energy, scale, and coherence, a key capability that essentially gives the quantum system greater capacity to address longer and more complex algorithms. The latest 1,200-qubit system includes both the new fabrication stack and new architecture. The higher coherence is a key reason why Advantage2 is more powerful than its predecessor.
“Higher coherence time means faster time to solution,” Baratz said. “We increased the connectivity. In the Advantage system, each qubit is connected to 15 others. Now with Advantage2, each qubit is connected to 20, which allows us to solve larger problems. We’ve increased the energy scale of processing by 40 percent, which means we can specify problems with higher precision and get higher quality solutions to the problems.”
Next up on the D-Wave roadmap is a 4,800-qubit Advantage2 QPU, with multiple samples currently under calibration and with a new fabrication stack that he said he expects will be available on Leap soon. The eventual goal is QPUs with more than 7,000 qubits, though it’s unclear when it will become available.
Baratz noted that with the 1,200-qubit Advantage2 prototype and the fast annealing protocol, the company was finally able to show the violation of the Bell inequality, which he called the “standard signature of a quantum computer.” It indicates that particles in the system are entangled and showing correlations beyond what a classical computer offer.
“We’ve been able to show superposition and we’ve been able to show entanglement,” Baratz said. “We’ve been able to show tunneling, we’ve been able to show coherent quantum annealing – all the things that make a quantum computer a quantum computer, and then allow us to use quantum mechanical effects to solve problems better than classical computing. But this one piece, which is an important piece, we haven’t been able to show. We’ve now really demonstrated it.”
D-Wave reached this milestone in part by enabling some qubits to act as processing qubits, while others are used to interrogate and read out data from the processing qubits while still in supposition, he said. Other qubits are used to force the compute qubits into ground states or excited states.
The company is working on its gate-model initiative, though Baratz gave no indication of when it will be released. He pointed to the synergy seen in developing annealing and gate-model systems. Both require high qubit coherence, so the investments D-Wave is making in materials and fabrication processes to improve coherence will apply to both. In addition, he noted that the ability to interrogate qubits while they are in superposition was developed as part of the gate-model program and ported to the annealing work.
Eventually there will be a time when enterprises can turn to D-Wave for both annealing and gate-model quantum computing.
“We know that our customers have problems that cannot be solved with an annealing quantum computer and will require gate-model system,” Baratz said. “We want to be able to provide a full suite of quantum technologies to solve their problems and, frankly, we’ll be the only one because we’re the only company working with the annealing piece.”
The 2ⁿᵈ figure tells the whole story in my mind, as a way to optimize revenue, with a single quantum traveling salesperson, tunneling coherently through all cities to visit, simultaneously, via superposed quantum teleportation, enhanced by entanglement. Add a pinch of time-reversed cat qubits to this mix and whamo! Voiding Bell’s inequality, the phone rings *after* you’ve picked it up, to answer the constant flow of error-corrected sales calls. It’s like quantum chromodynamics for the wallet! 8^b