IBM today, for the first time, published its road map for the future of its quantum computing hardware. There is a lot to digest here, but the most important news in the short term is that the company believes it is on its way to building a quantum processor with more than 1,000 qubits — and somewhere between 10 and 50 logical qubits — by the end of 2023.
Currently, the company’s quantum processors top out at 65 qubits. It plans to launch a 127-qubit processor next year and a 433-qubit machine in 2022. To get to this point, IBM is also building a completely new dilution refrigerator to house these larger chips, as well as the technology to connect multiple of these units to build a system akin to today’s multi-core architectures in classical chips.
IBM’s Dario Gil tells me that the company made a deliberate choice in announcing this road map and he likened it to the birth of the semiconductor industry.
“If you look at the difference of what it takes to build an industry as opposed to doing a project or doing scientific experiments and moving a field forward, we have had a philosophy that what we needed to do is to build a team that did three things well, in terms of cultures that have to come together. And that was a culture of science, a culture of the road map, and a culture of agile,” Gil said.
He argues that to reach the ultimate goal of the quantum industry, that is, to build a large-scale, fault-tolerant quantum computer, the company could’ve taken two different paths. The first would be more like the Apollo program, where everybody comes together, works on a problem for a decade and then all the different pieces come together for this one breakthrough moment.
“A different philosophy is to say, ‘what can you do today’ and put the capability out,” he said. “And then have user-driven feedback, which is a culture of agile, as a mechanism to continue to deliver to a community and build a community that way, and you got to lay out a road map of progress. We are firm believers in this latter model. And that in parallel, you got to do the science, the road map and the feedback and putting things out.”
But he also argues that we’ve now reached a new moment in the quantum industry. “We’ve gotten to the point where there is enough aggregate investment going on, that is really important to start having coordination mechanisms and signaling mechanisms so that we’re not grossly misallocating resources and we allow everybody to do their piece.”
He likens it to the early days of the semiconductor industry, where everybody was doing everything, but over time, an ecosystem of third-party vendors sprung up. Today, when companies introduce new technologies like UV lithography, the kind of road maps that IBM believes it is laying out for the quantum industry today help every coordinate their efforts.
He also argues that the industry has gotten to the point where the degree of complexity has increased so much that individual players can’t do everything themselves anymore. In turn, that means various players in the ecosystem can now focus on specializing and figuring out what they are best at.
“You’re gonna do that, you need materials? The position technology? Then in that, you need the device expertise. How do you do the coupling? How do you do the packaging? How do you do the wiring? How do you do the amplifiers, the cryogenics, room temperature electronics, then the entire software stack from bottom to top? And on and on and on. So you can take the approach of saying, ‘well, you know, we’re going to do it all.’ Okay, fine, at the beginning, you need to do all to integrate, but over time, it’s like, should we be in the business of doing coaxial cabling?”
We’re already seeing some of that today, with the recent collaboration between Q-CTRL and Quantum Machines, for example.
Gil believes that 2023 will be an inflection point in the industry, with the road to the 1,121-qubit machine driving improvements across the stack. The most important — and ambitious — of these performance improvements that IBM is trying to execute on is bringing down the error rate from about 1% today to something closer to 0.0001%. But looking at the trajectory of where its machines were just a few years ago, that’s the number the line is pointing toward.
But that’s only part of the problem. As Gil noted, “as you get richer and more sophisticated with this technology, every layer of the stack of innovation ends up becoming almost like an infinite field.” That’s true for the semiconductor industry and maybe even more so for quantum. And as these chips become more sophisticated, they also become larger — and that means that even the 10-foot fridge IBM is building right now won’t be able to hold more than maybe a million qubits. At that point, you have to build the interconnects between these chambers (because when cooling one chamber alone takes almost 14 days, you can’t really experiment and iterate at any appreciable speed). Building that kind of “quantum intranet,” as Gil calls it, is anything but trivial, but will be key to building larger, interconnected machines. And that’s just one of the many areas where inventions are still required — and it may still take a decade before these systems are working as expected.
“We are pursuing all of these fronts in parallel,” Gil said. “We’re doing investments with horizons where the device and the capability is going to come a decade from now […], because when you have this problem and you only start then, you’ll never get there.”
While the company — and its competitors — work to build the hardware, there are also plenty of efforts in building the software stack for quantum computing. One thing Gil stressed here is that now is the time to start thinking about quantum algorithms and quantum circuits, even if today, they still perform worse on quantum computers than classical machines. Indeed, Gil wants developers to think less about qubits than circuits.
“When [developers] call a function and now it goes to the cloud, what is going to happen behind the scenes? There are going to be libraries of quantum circuits and there’s going to be a tremendous amount of innovation and creativity and intellectual property on these circuits,” explained Gil. And then, those circuits have to be mapped to the right quantum hardware and indeed, it looks like IBM’s vision here isn’t for a single kind of quantum processor but ones that have different layouts and topologies.
“We are already, ourselves, running over a billion quantum circuits a day from the external world — over a billion a day,” Gil said. “The future is going to be where trillions of quantum circuits are being executed every day on quantum hardware behind the scenes through these cloud-enabled services embedded in software applications. ”