IBM Is Planning to Build Its First Fault-Tolerant Quantum Computer by 2029

This week, IBM announced a pair of shiny new quantum computers.

The company’s Condor processor is the first quantum chip of its kind with over 1,000 qubits, a feat that would have made big headlines just a few years ago. But earlier this year, a startup, Atom Computing, unveiled a 1,180-qubit quantum computer using a different approach. And although IBM says Condor demonstrates it can reliably produce high-quality qubits at scale, it’ll likely be the largest single chip the company makes until sometime next decade.

Instead of growing the number of qubits crammed onto each chip, IBM will focus on getting the most out of the qubits it has. In this respect, the second chip announced, Heron, is the future.

Though Heron has fewer qubits than Condor—just 133—it’s significantly faster and less error-prone. The company plans to combine several of these smaller chips into increasingly more powerful systems, a bit like the multicore processors powering smartphones. The first of these, System Two, also announced this week, contains three linked Condor chips.

IBM also updated its quantum roadmap, a timeline of key engineering milestones, through 2033. Notably, the company is aiming to complete a fault-tolerant quantum computer by 2029. The machine won’t be large enough to run complex quantum algorithms, like the one expected to one day break standard encryption. Still, it’s a bold promise.

Quantum Correction

Practical quantum computers will be able to tackle problems that can’t be solved using classical computers. But today’s systems are far too small and error-ridden to realize that dream. To get there, engineers are working on a solution called error-correction.

A qubit is the fundamental unit of a quantum computer. In your laptop, the basic unit of information is a 1 or 0 represented by a transistor that’s either on or off. In a quantum computer, the unit of information is 1, 0, or—thanks to quantum weirdness—some combination of the two. The physical component can be an atom, electron, or tiny superconducting loop of wire.

Opting for the latter, IBM makes its quantum computers by cooling loops of wire, or transmons, to temperatures near absolute zero and placing them into quantum states. Here’s the problem. Qubits are incredibly fragile, easily falling out of these quantum states throughout a calculation. This introduces errors that make today’s machines unreliable.

One way to solve this problem is to minimize errors. IBM’s made progress here. Heron uses some new hardware to significantly speed up how quickly the system places pairs of qubits into quantum states—an operation known as a “gate”—limiting the number of errors that crop up and spread to neighboring qubits (researchers call this “crosstalk”).

“It’s a beautiful device,” Gambetta told Ars Technica. “It’s five times better than the previous devices, the errors are way less, [and] crosstalk can’t really be measured.”

But you can’t totally eliminate errors. In the future, redundancy will also be key.

By spreading information between a group of qubits, you can reduce the impact of any one error and also check for and correct errors in the group. Because it takes multiple physical qubits to form one of these error-corrected “logical qubits,” you need an awful lot of them to complete useful calculations. This is why scale matters.

Software can also help. IBM is already employing a technique called error mitigation, announced earlier this year, in which it simulates likely errors and subtracts them from calculations. They’ve also identified a method of error-correction that reduces the number of physical qubits in a logical qubit by nearly an order of magnitude. But all this will require advanced forms of connectivity between qubits, which could be the biggest challenge ahead.

“You’re going to have to tie them together,” Dario Gil, senior vice president and director of research at IBM, told Reuters. “You’re going to have to do many of these things together to be practical about it. Because if not, it’s just a paper exercise.”

On the Road

Something that makes IBM unique in the industry is that it publishes a roadmap looking a decade into the future.

This may seem risky, but to date, they’ve stuck to it. Alongside the Condor and Heron news, IBM also posted an updated version of its roadmap.

Next year, they’ll release an upgraded version of Heron capable of 5,000 gate operations. After Heron comes Flamingo. They’ll link seven of these Flamingo chips into a single system with over 1,000 qubits. They also plan to grow Flamingo’s gate count by roughly 50 percent a year until it hits 15,000 in 2028. In parallel, the company will work on error-correction, beginning with memory, then moving on to communication and gates.

All this will culminate in a 200-qubit, fault-tolerant chip called Starling in 2029 and a leap in gate operations to 100 million. Starling will give way to the bigger Blue Jay in 2033.

Heisenberg’s Horse Race

Though it may be the most open about them, IBM isn’t alone in its ambitions.

Google is pursuing the same type of quantum computer and has been focused on error-correction over scaling for a few years. Then there are other kinds of quantum computers entirely—some use charged ions as qubits while others use photons, electrons, or like Atom Computing, neutral atoms. Each approach has its tradeoffs.

“When it comes down to it, there’s a simple set of metrics for you to compare the performance of the quantum processors,” Jerry Chow, director of quantum systems at IBM, told the Verge. “It’s scale: what number of qubits can you get to and build reliably? Quality: how long do those qubits live for you to perform operations and calculations on? And speed: how quickly can you actually run executions and problems through these quantum processors?”

Atom Computing favors neutral atoms because they’re identical—eliminating the possibility of manufacturing flaws—can be controlled wirelessly, and operate at room temperature. Chow agrees there are interesting things happening in the nuetral atom space but speed is a drawback. “It comes down to that speed,” he said. “Anytime you have these actual atomic items, either an ion or an atom, your clock rates end up hurting you.”

The truth is the race isn’t yet won, and won’t be for awhile yet. New advances or unforeseen challenges could rework the landscape. But Chow said the company’s confidence in its approach is what allows them to look ahead 10 years.

“And to me it’s more that there are going to be innovations within that are going to continue to compound over those 10 years, that might make it even more attractive as time goes on. And that’s just the nature of technology,” he said.

Image Credit: IBM

Jason Dorrier
Jason Dorrier
Jason is editorial director of Singularity Hub. He researched and wrote about finance and economics before moving on to science and technology. He's curious about pretty much everything, but especially loves learning about and sharing big ideas and advances in artificial intelligence, computing, robotics, biotech, neuroscience, and space.
Don't miss a trend
Get Hub delivered to your inbox