IBM plans a huge leap in superfast quantum computing by 2023

September 15, 2020, 12:39 PM UTC

Tucked in the back of a laboratory at the IBM Research facility less than an hour north of New York City is a hulking mass of stainless steel and aluminum that looks like a sci-fi teleportation machine.

I say as much to Dario Gil, who is leading a tour of the Yorktown, N.Y., site, as I clamber onto the rim of the machine for a photo opp. A coterie of masked engineers looks on like proud parents. The director of IBM Research considers the comment for a moment. Gil’s eyes glint, and he replies in his Spanish accent, “Well, it eez one!”

Gil isn’t joking: I’m standing on a quantum computer, or at least the exterior “clamshell” of one that IBM is building. A form of quantum teleportation will take place inside through the strange quantum mechanical properties of “superposition” and “entanglement” that make quantum computers hum—even if this device won’t zap my own body’s molecules to another planet.

Beam me up, Scotty. Fortune senior writer Robert Hackett stands inside the refrigerator casing for IBM’s planned 1 million-qubit quantum computer.

Quantum computers are the next evolution of today’s speediest supercomputers; the machines are now being developed by a range of tech giants, such as Google, Intel, and Honeywell, and startups, like Rigetti and IonQ. The machines are expected to solve presently intractable problems in science and business with applications spanning chemical modeling, hedge fund portfolio-strategizing, drug discovery, and A.I. advancements.

The 8,000-pound monstrosity around me—a humongous shotgun-shell of a device suspended on a square, metal frame—is empty now (save for my body), but before the decade is through IBM plans to install within it a quantum computer that contains 1 million so-called qubits, the quantum computing equivalent of classical computing’s bits. The work-in-progress represents the world’s most ambitious, concrete plan for a quantum computer now in construction—at least that’s publicly known.

An IBM engineer working on the refrigerator casing inside of which the company plans to install a 1 million-qubit quantum computer by 2030. Credit: Connie Zhou for IBM.
An IBM engineer working on the refrigerator casing inside of which the company plans to install a 1 million-qubit quantum computer by 2030.

Many scientists say that quantum computers of the sort IBM and Google are building will require at least 1,000 qubits—and possibly as many as a million qubits—before the devices become truly commercially useful. Since the machines are finicky and prone to error, most of the qubits will be devoted to correcting for those issues.

To date most companies have struggled to produce quantum processors bearing more than double digit-qubits with fidelity. Yet progress continues apace and advances are expected in the coming months.

IBM is feeling confident enough in its abilities that it is unveiling its full quantum hardware roadmap for the decade ahead. A week before unveiling its plans publicly on Tuesday, IBM gave Fortune a tour of its Thomas J. Watson Research Center, including a sneak peek at its upcoming machine.

IBM CEO Arvind Krishna (left) and director of IBM Research Dario Gil are pictured with a 10-foot-tall and six-foot-wide “super-fridge,” nicknamed “Goldeneye.” Credit: Connie Zhou for IBM

At the moment, quantum computers can’t do much beside run limited experiments—although that may soon change. IBM’s announcement comes almost a year after Google, a top rival, claimed to have achievedquantum supremacy,” a technical demonstration of the superiority of a quantum computer over a classical computer on an arcane task. (IBM disputes Google’s results, arguing that the IBM-built supercomputer at Oak Ridge National Laboratory in Tenn., called Summit, could have achieved a result faster than the Google team gave it credit for.)

While Google has kept relatively quiet since then, IBM is making more noise about its business plans. Big Blue aims to stake its claim on a nascent industry, corralling partners, and preparing everyone for an eventuality that may be coming sooner than anyone expects: the day quantum computers will become as useful for businesses as cloud computing, dominated by Amazon and Microsoft, is today. (Ever-struggling IBM, which named a new CEO earlier this year, places a distant fifth place on the cloud computing leaderboard.)

Despite IBM’s grand plans, a competitor, like Google, could still zoom ahead. Hartmut Neven, head of Google’s Quantum A.I. research group, told MIT Technology Review earlier this year that he believes Google is poised to build a 1 million-qubit quantum computer by 2030 too. (Google declined to comment on IBM’s news.)

But IBM is going a step further by cracking open its playbook and revealing its targets in detail. The company debuted a 65-qubit “hummingbird” processor this year. Next year the company plans to put out a quantum processor, called “eagle,” that contains 127 qubits. In 2022, it aims to introduce a processor that has 433 qubits, dubbed “osprey.”

IBM’s planned 1,121-qubit “Condor” quantum processor, slated for 2023, could be the company’s first truly useful quantum computer for businesses.

IBM’s roadmap hinges on a milestone slated for 2023. The company plans to release a 1,121-qubit quantum processor, called “condor.” Gil says it believes the moment will mark an “inflection point” when businesses will finally be able to do certain computations quicker, cheaper and more efficiently on a quantum computer versus on an ordinary computer; he calls this the dawning era of “quantum advantage.”

“Building that machine itself will be an incredible accomplishment, but it will tell us with a lot of proof whether that other one”—the 1 million-qubit machine—”is possible,” Gil says.

If IBM can meet the 1,121-qubit goal in the next three years, the company says it is confident it will make its 2030 deadline for a 1 million-qubit machine. That computer will be the culmination of decades of scientific research and breakthroughs, including foundational work done by Charles Bennett, a pioneering scientist and IBM fellow.

“We know that thing on the right is a world changing machine,” Gil says, referring to the 1 million-qubit processor. “It’s on the short list of seminal events in the history of information technology. To build a machine like that is the grand aspiration of the field.”

Along the way, IBM says it expects to achieve exponentially improving “quantum volume,” an invented measure that factors in a processor’s propensity for operational errors. As the machines become more controllable and less error-prone, they become more useful.

The Montreal quantum computer, highlighted in green on the right-side chart, refers to IBM’s already released 27-qubit “falcon” processor.

Companies like JPMorgan Chase and ExxonMobil are already signed on as customers of IBM’s quantum hardware.

Scott Aaronson, a computer science professor and quantum theoretician at the University of Texas at Austin, says IBM’s plans are “ambitious,” though they fall in line with the timelines put forward by rival teams. “We don’t exactly have a shortage of plans and roadmaps in this hype-soaked field,” he says.

Aaronson says the key will be how IBM “keep[s] down the error rate” of its quantum computers, since the qubit technology can be “a pain to integrate and calibrate.”

“Events haven’t been kind either to those who’ve promised useful devices right around the corner, or to those who said this was fundamentally impossible and all a fantasy,” Aaronson adds.

An IBM engineer tinkering on one of the company’s quantum computers. Credit: Connie Zhou for IBM

The competition is a tight one. This summer Honeywell revealed a quantum computer that it claimed was more powerful than any other, as measured by an IBM’s own preferred quantum volume metric. Quantum volume grades a machine based on three characteristics: the number of qubits, the level of connectivity between the qubits, and the qubit’s general quality, meaning their ability to avoid error.

It’s anyone’s race though. In August, IBM released a 32-qubit quantum processor with a quantum volume of 64, matching Honeywell. The competition is expected to remain neck-and-neck for the foreseeable future.

IBM and Honeywell are building different types of hardware. Honeywell relies on vacuum-suspended Ytterbium atoms (the “ion trap” way), whereas IBM uses super-cooled semiconductors (“superconducting qubits”). The methods have unique pros and cons, and it’s unclear which approach may ultimately take the lead.

On a broader scale, a geopolitical contest is unfolding in which the U.S. and Europe is vying with China for dominance in the nascent industry. This month the White House pledged $1 billion to the effort alongside companies including IBM, Microsoft, and Intel. Meanwhile, China is spending billions of dollars of its own and publishing groundbreaking experiments headed by its leading scientific light, the quantum physicist Pan Jianwei.

To the victor will go tremendous spoils. Beyond untold advances in science, technology, and business, there are serious potential ramifications for national security. Researchers are already working to shore up the foundations of data security, which could crumble at the hands of a fully functional quantum computer in time to come.

The IBM Research Center represents the company’s earliest site to reopen amid the pandemic’s shutdowns. The engineers are working hard to meet their deadlines, and this reporter even caught Gil pulling aside some engineers to tell them to remember to take time off.

All this is a marathon, after all, not a sprint.

Read More

Artificial IntelligenceCryptocurrencyMetaverseCybersecurityTech Forward