D-Wave revives 'quantum supremacy' claims for new Advantage2 computer

Quantum computing pioneer D-Wave Quantum on Tuesday announced the general availability of its sixth-generation quantum computer, the Advantage2. The company said the Advantage2 offers orders-of-magnitude greater performance compared to its prior system, expanding the tasks the company can accomplish in optimization problems.

The machine even achieves the long-sought goal of quantum “supremacy,” says the company, despite that term’s highly controversial past.

Also: If you’re not working on quantum-safe encryption now, it’s already too late

“This is a really historic moment for both D-Wave and the quantum computing industry,” said D-Wave CEO Alan Baratz in an interview via Zoom. “Fundamentally, our technology is doing something that can’t be touched classically.”

“We have customers running in production, getting value from our technology today. And everything we’ve done is to make it easy to use, so our customers have high uptime, and it can be very responsive to their needs. This is fundamentally differentiating us from the rest of the industry.”

“It’s fundamentally a massive leap in performance,” said Trevor Lanting, chief development officer at D-Wave, in the same interview.

The Advantage2 improves upon its predecessor along multiple vectors. Most prominent, by making a total of 20 connections from each of 4,400 qubits to other qubits, up from 15 connections in the original Advantage, the company is able to double the “coherence time,” the narrow window in which quantum calculations can be measured to get an output before the quantum state collapses. 

Also: Google’s quantum breakthrough is ‘truly remarkable’ – but there’s more to do

“Doubling the coherence gets answers much faster,” noted Lanting, on the order of 10,000 times faster than before. That may sound contradictory, but longer coherence leads to greater accuracy, and so it’s all about how much time is required to get to an answer of acceptable quality. 

Two kinds of quantum machines

D-Wave stands out from most other quantum computing projects in a big way, including machines from competitors IonQ and Rigetti Computing. 

In quantum computing, there are two rival technology approaches. 

One approach is called the “gate” model, where IonQ and others use a variety of exotic technologies such as trapped ions to fashion representations of ones and zeros that display quantum-mechanical properties such as entanglement and tunneling. 

The gate model seeks to create a universal quantum computer that can simulate any of the classical computing circuits, such as an “AND” circuit or an “XOR” circuit, out of which any legitimate operation can be fashioned. That makes such machines immediately applicable to problems such as solving differential equations.

The hypothesis is that the gift of nature, quantumness, allows such traditional circuits to operate in a highly parallel fashion, vastly speeding up computation.

Also: Amazon credits analog with making Ocelot a more-efficient quantum chip

Since its founding in 1999, D-Wave has pursued a second, far less fashionable, approach called “quantum annealing.”

The metallurgical art of annealing, derived from the verb “to burn” in Old German, involves heating a substance, such as metal, and then allowing it to gradually cool, forming a new, more resilient form. 

An algorithm on a computer simulates the metallurgical process: It allows the variables of a problem to figuratively heat up, by loosening constraints on what values they can take, so that more possible configurations can be tried. Through such trial and error, values of a problem eventually settle into a stable, low-energy state that is the solution to a problem.  

D-Wave has built a quantum version. Like a traditional computer chip, the D-Wave processor is made up of various metal layers arranged in a matrix of computing cores that connect to one another over a communications mesh. 

The design is fabricated for D-Wave under contract with SkyWater Technology of Bloomington, Minn.

The annealing process doesn’t lend itself to the universal operations of the gate computers, but it makes D-Wave’s machines very good at one type of problem: optimization. Optimization involves multiple variables that have to coalesce into an ideal state to solve a problem, such as resource scheduling or supply chain management. 

The Advantage2 broadens the scope of problems the annealing technology can solve by juggling more variables, said Baratz.

“Today, we can solve problems that are in the hundreds to thousands to tens of thousands of variables range,” he said. “We ultimately need to be solving problems that are in the tens of millions of variables range — this starts to open that up for us.”

Currently, D-Wave addresses one quarter of the market for quantum computing, the optimization problems — a capability unique to the company’s annealing technology. Two other broad categories that the machine can ultimately address are linear algebra and factorization problems, while the fourth, differential equations, is limited to gate-style machines. 

“So, we can address roughly three-quarters of the total addressable market, one-quarter of that exclusively ours.” In 20 years, just the optimization portion may be worth $200 billion annually, he said.

Open for business

The general availability of Advantage2 follows years of testing on the prototype version of the machine by some D-Wave customers, including running over 20 million customer jobs on the machine, said Lanting.

Twenty-five of the top 2,000 global corporations, including MasterCard and Japanese telecom giant NTT-Docomo, are already using the machine, said Baratz.

Also: Microsoft’s quantum chip Majorana 1 is a few qubits short

Customers gain access via D-Wave’s Leap cloud computing service. The cost is $25,000 per quarter for two developer seat licenses, with what Baratz said is “essentially unlimited time” on the computer. For production runs, “we charge by the application,” he said, “based on the characteristics of the application, how complex the application is,” with costs ranging from a “few hundred thousand dollars a year to a few million dollars a year.”

The optimization problems are highly relevant, Baratz told me, amidst supply chain issues caused by the US’s re-ordering of global tariffs and trade.

“These are challenging times for businesses,” he said. “Companies need to operate with greater efficiency in order to hold the bottom line in light of all these challenges.”

A lot of those challenges, said Barataz, “can be addressed by solving hard computational problems, whether it’s figuring out how to optimize your supply chain or figuring out how to improve operating efficiency in the business. At the core of a lot of these issues is solving hard optimization problems. And that’s what we’re all about.”

Baratz said that in an effort to ensure the Leap service is “always on the latest generation,” all machines in the cloud will be upgraded over time to Advantage2.

Also: This quiet AI upgrade actually changed my life

D-Wave also sells a complete system for those who can afford the multi-million-dollar price tag. It’s important to do both, said Baratz. “For our government and research customers that are looking at exploring new kinds of workloads, and tighter integration with supercomputers, they need to own the system,” he said.

The company is working on how to train generative AI models, such as large language models. “You could envision hyperscalers buying these systems to be a part of their AI infrastructure as well,” he said of Google and other tech giants.

Claims of supremacy

Advantage2 already has some intriguing scientific data to back it up. In an article published in the April issue of the scholarly journal Science, D-Wave researchers describe using Advantage2 to solve a complex problem in materials physics: measuring a magnetic material and extracting its properties as it undergoes what’s called a phase transition.

The Advantage2, in a prototype version running only 1,200 qubits, was compared against the Frontier supercomputer running at the Oakridge National Laboratories of the US Department of Energy.

By one measure, it takes the Advantage2 less than a dollar of electricity to compute the solution to the problem versus what would be “the global annual output [in electrical consumption] for classical computers,” said Lanting.

Also: Google’s upcoming AI smart glasses may finally convince me to switch to a pair full-time

“This is giving you tools to do calculations that not only are beyond the reach of classical supercomputers, but are really, fundamentally, much more energy-efficient than any other means of doing these hard calculations,” said Lanting. 

Baratz refers to the article as “the supremacy paper,” which may be a source of controversy for some in the field. 

The paper itself doesn’t use the term “quantum supremacy,” which physicist John Preskill defined in 2012 to mean results that “vastly exceed” those of traditional computers. 

Instead, the Science article relies on the phrase “Beyond-classical computation.” The paper was co-authored by numerous researchers from institutions around the world, noted Baratz, which meant that D-Wave did not have unilateral control over the language.

However, for his part, Baratz is convinced there really is a supremacy demonstrated by the system that agrees with Preskill’s definition. 

“We do believe that what we demonstrated in the science paper is quantum supremacy as John Preskill defined it,” said Baratz. “The paper really showed that you can’t do this in any way via classical means.”

“During the pandemic, there were some folks that said we should move away from the use of the term supremacy,” said Baratz, referring to debates that maligned the use of “supremacy” in its connection with “white supremacy.”

Also: What is sparsity? DeepSeek AI’s secret, revealed by Apple researchers

“We’ve never bought into that because we believe that supremacy is the best defined term for what we have accomplished, and nobody else has done it yet.”

Another research area D-Wave is pursuing is what would be a “quantum proof of work” for blockchain crypto, a way to achieve the tasks of Bitcoin and Ether without using as much energy. 

“The idea would be to produce a more energy-efficient blockchain that harnesses this supremacy calculation,” said Baratz. 

A blockchain on quantum, by its nature, would be a market for computer sales, said Baratz, because the machines doing proof of work need to be distributed globally rather than run in a cloud service.

Lest you be concerned about quantum hacking the blockchain, the ability to overcome the 50% threshold of compute that maintains blockchain integrity, “right now, there isn’t a way to short circuit that with today’s quantum technology,” he said, though that could change in the future.

The competition is no competition

Baratz is unfazed by the recent raft of announcements from deep-pocketed tech giants Google, Amazon, and Microsoft.

“None of those three are competition to us currently because they’re all gate-model systems, and we’re annealing, and they are different applications.”

Still, Baratz is happy to opine on the relative strengths of the group. 

The Google Willow processor, announced in December, is fairly impressive, he says. “It was a very strong result,” he said. “They did demonstrate that it reduces errors,” and that the error reduction improves as the number of gate-style qubits increases.

Also: Most AI chatbots devour your user data – these are the worst offenders

Amazon’s Ocelot chip, announced in February, “was nowhere near as strong,” he said. 

As for Microsoft’s Majorana1 chip, which has generated controversy in the field, “it was a sham,” he said. “They did not demonstrate even one qubit, let alone eight.”

Baratz did not remark on IonQ or Rigetti, though in the past, he has said the optimization problem approach of annealing is far more applicable to current market demand than are those companies’ gate-style machines.

All three companies have relatively small revenue at present, so the quantum cohort is still in its early days.

A roadmap for quantum

Moving forward, the roadmap for D-Wave involves a couple of directions. The company is working on an Advantage3, which will take a new direction in combining multiple chips rather than relying on one chip. 

“Up until now, it’s been about more qubits per chip,” Baratz reflected. “With Advantage3, we’ll be interconnecting chips because we believe that will allow us to scale a lot faster. Think about it, 4,000 or 5,000 qubit chips [combined together] is 20,000 qubits overnight,” he said, while allowing that it only sets the bar higher for Lanting. “Trevor breaks out in a cold sweat because he has to deliver it.”

Also: Where the cloud goes from here: 8 trends to follow and what it could all cost

Other important additions include digital computations inside the analog-based annealing circuitry. “We’re now at the point where we’re actually able to perform some digital computations inside the analog annealing fabric,” explained Baratz, “which we think will open up some incremental use cases for us.”

At the same time, “we’re also building a gate-model system,” said Baratz. “The near-term goals for that are basically to demonstrate high-fidelity qubits and error correction for logical qubits.”

The time frames for Advantage3 and a gate-style computer are vague — “some time in the future,” said Baratz.

When pressed, he said, “I’d love to be able to show some proof points in each of those this year, but we still have a lot of work to do.”

Get the morning’s top stories in your inbox each day with our Tech Today newsletter.