SAN FRANCISCO, June 10 (Reuters) - International

Business Machines ( IBM ) on Tuesday said it plans to have a

practical quantum computer by 2029, and it laid out the detailed

steps the company will take to get there.

Quantum computers tap into quantum mechanics to solve

problems that would take classical computers thousands of years

or more. But existing quantum computers must dedicate so much of

their computing power to fixing errors that they are not, on

net, faster than classical computers.

IBM ( IBM ), which also said it aims to have a much larger system by

2033, plans to build the "Starling" quantum computer at a data

center under construction in Poughkeepsie, New York, and said it

will have about 200 logical qubits. Qubits are the fundamental

unit of quantum computing, and 200 qubits would be enough to

start showing advantages over classical computers.

IBM ( IBM ) is chasing quantum computing alongside other tech giants

such as Microsoft ( MSFT ), Alphabet's Google and

Amazon.com ( AMZN ), as well as a range of startups that have

raised hundreds of millions of dollars in capital.

All of them are tackling the same basic problem: Qubits are

fast but produce a lot of errors. Scientists can use some of a

machine's qubits to correct those errors, but need to have

enough left over for doing useful work.

IBM ( IBM ) changed its approach to that problem in 2019 and says it

believes it has landed on a new algorithm that will drastically

reduce the number of qubits needed in error correction.

In an interview, Jay Gambetta, the vice president in charge

of IBM's ( IBM ) quantum initiative, said the company's researchers took

a different tack than they had historically, when they would

work out the scientific theory of an error-correction method and

then try to build a chip to match that theory.

Instead, IBM's ( IBM ) quantum team looked at which chips were

practical to build and then came up with an error-correction

approach based on those chips. That has given IBM ( IBM ) confidence to

build a series of systems in between this year and 2027 that

will eventually result in larger systems.

"We've answered those science questions. You don't need a

miracle now," Gambetta said. "Now you need a grand challenge in

engineering. There's no reinvention of tools or anything like

that."