CNET también está disponible en español.

Ir a español

Don't show this again

Computers

Nuclear weapons lab buys D-Wave's next-gen quantum computer

Los Alamos National Laboratory will install D-Wave's 5,000-qubit Advantage system in 2020.

D-Wave is trying to bring quantum computing out of the research lab. This is one of its chips.

D-Wave is trying to bring quantum computing out of the research lab. This is one of its chips.

D-Wave

It's been a busy time in quantum computing: IBM announced its biggest quantum computer so far, and Google reportedly has cracked a problem with its quantum computer faster than a conventional computer could. Now D-Wave, a rival with a different approach to the technology, has announced that nuclear weapons research site Los Alamos National Laboratory (LANL) is the first customer for its next-generation machine.

Quantum computers are a totally different breed of machine than the classical computers that power everything from your smartphone to Amazon's vast e-commerce operation. Quantum computers rely on the weird physics of the very small to perform calculations classical machines  can't solve -- at least in principle. So far, the technology is very, very experimental.

If all goes as planned, though, quantum computers could open up new computing territory even as engineers struggle to wring more performance out of conventional machines.

Researchers at Google, IBM, Microsoft, Intel and Rigetti Computing are focusing on building universal quantum computers -- machines that can tackle any computing problem. D-Wave, though, has a different approach called annealing that's limited to a much narrower range of problems. For example, annealers can't tackle the most famous quantum computing party trick, the ability to crack today's encryption technology.

To be fair, universal quantum computers can't yet, either, because they don't have enough qubits -- the fundamental unit of quantum computing data storage and processing. And those qubits aren't stable enough, despite the fact that quantum computers run only a hair's breadth away from absolute zero, the coldest possible temperature. There's work afoot to pave over that instability with error correction technology, but that requires more qubits, so it's not an easy fix.

Toward universal quantum computing

Everybody, of course, is racing to improve their designs to get to a useful, universal quantum computer.

"We believe we'll get to universal before they get to error correction," said Alan Baratz, D-Wave's chief product officer, in an interview earlier this year.

D-Wave's next-gen Advantage quantum computer connects qubits with a new, higher-performance technology it calls Pegasus.

D-Wave

Naturally, IBM thinks its starting point is better.

"I'm not a believer in the annealing approach," Dario Gil, director of IBM Research, said in an August interview. Specifically, he's not convinced that an annealer would be able to outperform a well-designed classical computer.

IBM's newest Q quantum computer has 53 qubits, and Google's Bristlecone quantum computing chip has 72 qubits. D-Wave is up to 2,000 qubits today, but because its annealing approach is different, you shouldn't compare those numbers.

What you can compare is D-Wave's own qubit increase. Next year's Advantage line, the model LANL is buying, will reach 5,000 qubits. Plus they're interconnected in a cleverer way to squeeze more performance out of the machine, D-Wave said. 

D-Wave customers

But D-Wave has convinced some that its approach is useful -- including LANL, which already uses a D-Wave machine that it'll upgrade next year.

"This is the third time we will have upgraded our D-Wave system. Each upgrade has enabled new research into developing quantum algorithms and new tools in support of Los Alamos' national security mission," said Irene Qualters, LANL's associate director for simulation and computation, in a statement. "Quantum computing is a critical area of research for Los Alamos."

Another customer is financial services company CogniFrame, D-Wave said.

The company believes its machines are useful for a variety of other tasks besides algorithm development and financial modeling, though. On its list are airline scheduling, election modeling, quantum chemistry simulation, car design and shipping optimization.