Early quantum computing pioneer D-Wave again asserted that at least for D-Wave the commercial quantum era has begun. Speaking at its first in-person Analyst Day last week, CEO Alan Baratz touted the companys steady technical progress, assembled a few key customers to recount their D-Wave experience, and closed with a Q&A that provided details on its go-to-market approach and the costs to customers. The overall message of the day was a close reprise of Baratzs Qubit23 message last January D-Wave is open for commercial business.

Heres the Baratz pitch:

Many of you have heard me say this before, but its a very important point, D wave is absolutely unique in the quantum industry. Not only were we the first, but we are the only commercial quantum computing company. We have businesses that are either working with us today to develop applications to benefit their business operations, or they have moved those applications into production, and their businesses are benefiting today. Moreover, the quantum industries is absolutely at a watershed moment, said Baratz.

We are now at the point where quantum is transitioning from research experimentation to actual in production business support. But theres only one company that can say that and thats D-Wave because were the only company in the world that has quantum computers, hybrid solvers and quantum services that are able to support business applications in production. Everybody else in the quantum industry is still working on developing their systems. Thats a critically important point. That also means that D-Wave is the company that is single handedly building the quantum marketplace, because were the only ones that can do it.

These are broad claims, and many in the quantum industry disagree, even as widespread debate continues over when (and how) quantum computing will deliver value.

D-Wave, founded in 1999, has long championed a form of quantum computing quantum annealing (QA) that works quite differently from the gate-based approaches being pursued by the majority of quantum computer developers. The company also has a new gate development effort, but Baratz believes those systems are far from ready for everyone Ill say it again, we are at least seven years away from the gate model system being able to solve a real-world problem because theres just no evidence that a gate model system can solve a real-world problem without error correction.

The big bet D-Waves big bet is that quantum annealing, which does not require active error correction (EC), is the only current quantum option; its been shown to be effective for certain classes of problems, especially optimization. For two decades, D-Wave has been steadily improving the characteristics of its machines including coherence times and connectivity. Its systems support higher qubit counts because they dont need EC. The company has developed hybrid quantum solvers for a variety of application areas, and it now says it is verticalizing efforts around the long hanging fruit. Critically, it has a few customers using its systems for at least some production-environment jobs with broader roll-outs underway.

It also has pressure, not least from financial markets. In 2022, D-Wave went public (NYSE-QBTS) via the SPAC route, along with a handful of others, and twice faced delisting.

Theres a lot to unpack in the D-Wave journey. Is this D-Waves first-mover advantage moment, as Baratz declared, and the start of quantum annealing going mainstream while gate-based system continue development? Or something else?

Much of the material covered in the meeting wasnt new having become public over the course of the past couple of years. That said, it represents D-Waves self-view today, and while this was an Analyst Day intended to impress its audience, D-Wave has put many pieces to the QA puzzle together. Is it enough?

Before digging into a few meeting highlights (technology, use cases, go-to-market strategies and user costs) look at the three slides below (click on to enlarge) describing D-Wave position today. (Link to video of D-Waves Analyst Day)

Lets start with technology.

Unconstrained by active error correction, D-Wave systems have always had higher qubit counts. The first-gen Advantage system, now in use, has 5000 qubits. Key technology challenges have included increasing qubit connectivity, extending coherence times, and incorporating error mitigation. Advantage2, expected in 2023/24, will have 7000 qubits. By comparison, most gate based QPUs have far fewer (mostly single digits to 100ish). More qubits and connectivity mean larger, more complex problems can be mapped onto the processor. Better coherence means better answers.

Coincident with the analyst day, D-Wave announced it has calibrated a1,200+ qubit Advantage2 prototype, which will be available (Q1) in the companys Leap real-time quantum cloud service.

The new Advantage2 prototype features 1,200+ qubits and 10,000+ couplers, double the number of qubits and couplers over thepreviously released Advantage2 prototype. D-wave reported benchmarks demonstrate substantial advancements across a number of performance metrics compared to the Advantage quantum processing unit (QPU), including:

D-Wave says the new Advantage2 prototype is 20 times faster at solving spin glasses, an important family of classically hard optimization problem: Recent research has shown that compared to the Advantage system, the Advantage2 prototype grows quantum correlations twice as fast in materials simulation and shows significantly reduced errors in quantum simulation tasks. Further, it shows improved performance on constraint satisfaction problems, with the Advantage2 prototype beating the Advantage system 90% of the time.

An enhanced fabrication stack was developed to achieve the gains. Advantage2 system will mark the companys sixth-generation quantum system. Baratz noted the 1200-qubit Advantage2 prototype is already more powerful than our current 5000 qubit events. He also cited a recent Nature paper that showed while we are doing quantum annealing, we get a significant speedup over classical systems on hard optimization problems, in particular a polynomial speed up, not an exponential speed up.

The companys relatively new gate-based development program is also proceeding. D-Wave has chosen to work with fluxonium qubits and recently reported, manufactured and tested fluxonium qubits in a 2-dimensional circuit geometry. The measured coherence properties, with relaxation times in excess of 100 microseconds, are comparable to the current state-of-the-art for such qubits. In addition, the measured effective temperature of its fluxonium, 18 millikelvin, is among the best that has been reported in the scientific literature to date for superconducting qubits.

Baratz said D-Wave has been able leverage much of the technology developed for quantum annealing for its gate-based efforts, including for example, programming and readout able to achieve readout 20 times faster than anything thats ever been shown on gate model systems in the edge.

Theyre all using something called dispersive readout, which is not a great technology. We have already developed much better technology for doing that annealing, and its directly applicable to gate, and weve been able to show that, according to Baratz. The next step, he said, is to bring together the flexibility and high coherence of fluxonium qubits with the control capabilities that we developed to build a logical gate. Weve already designed the mask for those next step is to fabricate and test.

Turning quantum technology into practical solutions, deployed in a production environment, and, of course delivering sufficiently superior performance to classical systems to warrant the cost and effort is everyones goal. The projects reviewed by D-Wave customers covered a range of optimization activities are at various stages of roll-out: Pattison Food Group (E-commerce driver and delivery scheduling); Davidson Technologies (radar scheduling); IPG (tour scheduling); Vinci Energies (HVAC design).

Perhaps the furthest along is work by food products giant Pattison, presented by Lindsay Dukowski, senior manager, delivery scheduling. What began as a workforce scheduling project in 2020 was briefly paused by the Pandemic. It morphed into a successful ecommerce auto delivery scheduling application, now in use.

Of the initial project, Dukowski said, Across 13 different collective bargaining agreements, the rules to create a schedule were extremely complicated, and labor intensive, as well. So, we were looking at Workday (ERP solution), for instance, a leading application and even they couldnt solve this problem. I was actually leading that project at the time. We decided to partner with D-Wave to try to solve that problem.

Think scheduling how many cashiers you need, how many people in a bakery or the deli, and all of the scheduling rules and constraints, she said, including all the labor rules such as whos available, whos qualified to work in every department, and years of experience.

It takes between 8-to-12 hours a week per store to just complete. Once thats done, we put it into our workforce management application. Because these schedules are usually generated about three weeks in advance, theres a lot of last minute changes and edit sick time, new hires, terminations that have to be taken into consideration. All of that maintenance takes another 8-to-12 hours. If you think about 300 stores, its actually pretty expensive for us.

Pattison did a POC in a non-unionized store in July 2020 that was successful. Our leadership [said], youve proven you can solve this problem. Lets roll it out in the other stores (unionized) to see if you can actually really do this in a more complex environment.

COVIDs arrival changed the plan. You can imagine there are stores we had a lot of people calling in sick, a lot of people that werent willing to come in because they had family members that were at risk. COVID was crazy for the grocery industry, for every industry, but for grocery especially. We decided to kind of put that on pause, recalled Dukowski.

Pattison had e-commerce (order online for delivery) in about ~20-30 percent of its stores was pushing to expand that to all its stores. We decided to pivot and solve the e-commerce driver auto scheduling; so a very similar problem to coming up with the schedule at retail where youve got demand for drivers to deliver to multiple locations at the same time.

That project became QEDA Quantum Ecommerce Driver Automated scheduling. She recalls that e-commerce was in ~100 stores and 3-4 people that were manually creating these schedules every week and it took about 80 hours.

We went through the labor agreements. My team actually read through all of the documents. Then we met with the schedulers to find out the requirements. We converted those requirements to math equations, we built data pipelines to pull in all the necessary data from the source systems, converted that to an optimization code. We use the hybrid solver from D-Wave, did some code debugging and parameter tuning, outputted some schedules and then just kind of iterated the process until till it worked, Dukowski said.

The weekly manual effort for scheduling creation was reduced by 80% from 80 hours to 15. And that time is really for the maintenance and the edits that need to be made. Its difficult. We didnt try to solve that problem with this, it was just lets generate a schedule. Initially, the runtime per schedule is about two minutes, and then we generate 42 schedules each week. So, if you think about the timeline, we started in April, we did a pilot in August, and moved into production, October 2022.

I believe that at that time, we became the first the first company in North America to go live with a production system using quantum, said Dukowski.

Pattison has since returned to the workforce management application. We took the demand, we automated all of those scheduling rules, using the same project methodology and essentially the same tech stack to build a new model and pull in additional data. So, we have some new data pipelines we had to build, but leverage the work that we did with the E commerce driver scheduling, and were able to eliminate that eight to 12 hours, she said.

The scheduling challenge is one of those application areas D-Wave is currently focusing what Baratz called the low hanging fruit because the tools, hybrid solvers, and D-Wave quantum computer can effectively handle them. We are targeting manufacturing, logistics, and doubling down on key use cases such as workforce scheduling, resource allocation, and vehicle routing he said.

These projects remain non-trivial and take time, as indicated by the Pattison example. Baratz provided some detail around customer engagements.

Over the course of the last year and a half to two years since we launched our professional services organization, weve now done over 25 proofs of concept with customers. Were now at the point where they are starting to move into production [and] we expect a few more over the course of the next three months. This is a really important transition for us because of our business model. Ultimately, we want to focus on quantum computing as a service, because that allows us to build recurring revenue. And that puts us in a much more predictable revenue growth position going forward, said Baratz.

In Q&A, Baratz was asked to distinguish and contrast the overall 70 commercial customers with the ~25 POC cited earlier.

Keep in mind that not all customers [that] engage us do POC. We have a lot of customers that we affectionately call DIY customers, right? They come in, they buy in the old days of time and in new days a developer seat and then try to build the application themselves. What we know is that the customers that try to do it, themselves have a much lower probability of success than when they engage us in a POC, said Baratz.

Professional services helps them to figure out how to properly leverage the quantum computer, how to properly map the application, but as we go through it we can help them understand how to do so it doesnt always have to be us, he said. A lot of those 70 customers are those DIY customers. One of the things we are focused on is going back to them, and trying to move them to a POC, so that we can help them be more successful.

The standard pricing for proof of concept is about $350,000, said Baratz. We also have a $70,000, demonstrator engagement thats a one month engagement with a demonstration thats really quick and dirty to kind of feel out the solution to your problem. You cant run it because not linked to your environment. But you give us some data. We figure out how to solve a problem on your data, leveraging quantum hybrid solvers, said Baratz.

Moving into production is a deeper engagement. If the customer wants our help in trying to move to production, thats typically a custom agreement.

D-Wave has two production offerings.

Asked who owns the application developed with POCs and if D-Wave is able add newly-developed applications to its library, Baratz said, Wed love to have them, but our customers would not sometimes. Basically, what weve done is weve defined an interface, and typically what we say is anything thats developed below that interface we own. For example, if we make a change to a hybrid solver to support your application we own that; anything above that interface you own so typically, if the application is owned by our customers. But not always, sometimes weve been able to retain that.

As outlined by Baratz, D-Wave clearly has big ambitions and believes the timing is right. He notes that quantum annealing, sometimes disparaged in the past, has recently got a boost from government with favorable language in recent National Defense Authorization Act Weve already started getting calls from various defense agencies saying we need to learn more. The last call we got was from the army engineering research lab, he said.

Stay tuned.

More here:

Eyes on the Quantum Prize D-Wave Says its Time is Now - HPCwire