Honeywell is the quantum computing champ.
On Thursday, Honeywell said it boosted the performance of its quantum computer enough to best that of IBM's by one closely watched measure. The achievement gives Honeywell's machine the title of world's fastest quantum computer, at least for the time being.
The feat fulfills a promise Honeywell made in March and is a milestone in the company's ambitious effort to increase its quantum computing power by a factor of 10 each year for the next five years.
Honeywell's quantum computer, dubbed H0, scored 64 on a measure known as quantum volume. That test gauges both the total count of a machine's qubits -- the fundamental elements that process data in a quantum computer -- and how well the machine uses those qubits. The result bested the 32 IBM's Raleigh machine scored in January.
Quantum computing relies on the weird rules of atomic-scale physics to solve problems that are practically impossible for ordinary "classical" computers. At least that's the promise. So far quantum computers are mostly finicky research projects, only outperforming classical computers in one narrow "quantum supremacy" test at Google.
But as researchers add more qubits and learn to protect them from disturbances that derail calculations, expect quantum computers to become more powerful. Tasks they're expected to do well at include optimizing financial portfolios, machine learning, and designing new materials like solar panels or batteries.
Fresh start with quantum computing
Decades ago, Honeywell was a top maker of massive mainframes, but it sold the business and left the computing industry to other players. In recent years, it quietly assembled a team of 120 researchers in Colorado and Minnesota and used its expertise in materials science and industrial operations to tackle quantum computing.
Honeywell's current design, H0, will reach all the way from today's quantum volume of 64 to 640,000 as the company adds more qubits, said Tony Uttley, president of Honeywell's quantum computing division. "Think of an auditorium with a lot of seats. We've built the infrastructure to be an auditorium. We're filling it a few seats at a time," he said.
Another system, H1, is in development to reach beyond that, he added.
The number of qubits in a quantum computer is important. More qubits mean researchers can tackle more complex problems with more possible solutions, potentially having enough that groups of them can be ganged together into single virtual qubits that'll enable much more sophisticated computations. But qubit count on its own is an imperfect performance measure, which is why IBM came up with quantum volume.
"IBM's quantum volume benchmark is one of the earliest attempts to address this problem, and others will inevitably come down the pike as quantum computers become more capable," said 451 Research analyst James Sanders. But for now, it's easy to at least compare IBM and Honeywell machines.
IBM, now with 18 quantum computers, has made its own promise to boost performance with a plan to double quantum volume annually. IBM is confident in its own approach, said IBM Research leader Dario Gil, but says Honeywell's has potential, too. "It's a great thing that Honeywell is pushing an ambitious roadmap," Gil said last month.
Honeywell's quantum computers are made with a design called an ion trap that houses qubits. IBM and Google use a different approach, superconducting qubits, that requires the quantum computer be chilled to within a hair's breadth of absolute zero. Microsoft and Intel are pursuing other approaches they hope will let them leapfrog competitors.
It's a lot more variety than what you see with conventional computers. And it isn't clear if one approach will win out.
It might be that different aspects of a computing problem are farmed out to different types of quantum computers, Honeywell's Uttley said. What is sure is that quantum computers will run side by side with classical computers, boosting their abilities and not replacing them.