That lab, owned by IBM, is in pole position in a new race between Big Blue, Google, Microsoft and a horde of startups. Their goal? “Quantum advantage”—the point at which quantum computers exceed the abilities of the best conventional ones. Yes, good old IBM is holding its own with the trillion-dollar competition.
IBM hasn’t been associated with breakthrough innovation since its Watson AI won “Jeopardy!” in 2011. But quantum computing, which could see a breakthrough to commercialization by 2030, gives the 114-year-old stalwart of business computing a chance to reclaim some of its past glory.
It is working on larger clusters of quantum chips that it expects will enable large-scale computing in the next five years. In just the past month, the company announced a partnership with chip maker AMD to develop “quantum-centric supercomputers” and an update to its program to certify quantum developers.
Meanwhile, a deal frenzy among quantum computing startups saw nearly $2 billion injected into the field.
Models of the multitiered chandeliers inside IBM Quantum System Two.
Bringing ‘big iron’ back
The stakes are high. Quantum computers have the potential to defeat the encryption on which everything from passwords to bitcoin depends. They could also speed up the solution of the optimization problems that vex logistics companies and militaries alike. Eventually, owing to their ability to simulate matter at the subatomic scale, boosters say quantum computers could lead to breakthroughs in every area of materials science, medicine and agriculture.
They work by leveraging the quantum characteristics of matter at the atomic level. In theory, they can do some kinds of calculations billions of times faster than conventional computers.
The process is so counterintuitive it feels like magic. A slightly longer explanation: Unlike a classical bit on a chip, which can only be a 1 or a 0, a quantum bit—or “qubit”—can exist in a superposition of many states at once. This allows the computer to explore many potential solutions to certain problems almost instantaneously. One theory holds that this works because of parallel universes.
The multiverse aside, companies are already making hardware based on this science. IBM and Google are taking similar approaches and are experiencing the same basic issues: Qubits are so sensitive to disturbance that their operation can be thrown off by a stray cosmic ray—or an earthquake on the other side of the world.
Making qubits resistant to such disturbances—“fault tolerance”—is what Jerry Chow has been up to at IBM since 2010. A Harvard-educated physicist, Chow joined the company right after he finished a Ph.D. and now leads its quantum research efforts.
“We had a lot of good memories in here,” he says as he shows me the cramped and musty room in which he built one of the company’s earliest experimental quantum computers. Now, Chow presides over a warren of labs, filled to bursting with machines.
At the entrance to the company’s Thomas J. Watson Research Center—which was designed by Eero Saarinen and could double as a golden-age Bond villain lair—there stands a finished version of Chow’s latest masterpiece.
Its liquid helium cooling system humming quietly, its facade like a chromed version of the monolith from “2001,” IBM’s Quantum System Two churns away, working on real problems for early pioneers of the field. It’s the first of what the company hopes will someday be countless behemoths, a return to a storied past in which IBM delivered “big iron”—physically imposing computers whose scale reflected their power.
IBM has some advantages, including its own microchip fab—something only a few U.S. companies still have. It can continuously evolve its chips in-house alongside the hardware that supports them, like the intricate chandeliers of copper and wiring that cool its processors, says Chow.
While a true commercial market may still be a half-decade away, existing computers like Quantum System Two are already viable for serious research. There is also a growing cohort of engineers learning how to program these systems, who have collectively already published more than 3,000 papers using IBM’s hardware.
The road to 2030
By this decade’s end, IBM and its closest competitors should start generating significant revenue from quantum computing, says Mark Horvath, a vice president and analyst at Gartner.
Arvind Krishna, IBM’s chief executive, told me in a recent interview for WSJ’s Bold Names podcast he expects that to be a pivotal moment for quantum hardware makers—not unlike Nvidia’s star turn when generative AI took off.
But getting to that point requires overcoming plenty of engineering roadblocks. IBM is projecting it will soon supplant its current Heron quantum chips with a new generation, tied together into ever-larger clusters. The company expects to unveil its first large-scale, fault-tolerant quantum computer in 2029.
Google, which has its own road map, has said it is on the second stage of a six-stage journey toward releasing a robust and useful quantum computer. The head of Google’s AI efforts has said he is optimistic that within five years, people will be doing things with quantum computers that are impossible on conventional ones.
Earlier this week, Google announced it has been selected for the Defense Advanced Research Projects Agency effort to evaluate the claims of quantum computer makers. Darpa’s goal is to determine which of the current efforts can lead to a “utility-scale, fault-tolerant quantum computer”—in other words, more than just a science experiment—by 2033.
Because Google and IBM are taking a similar approach, they face similar problems. Among them: How to cool their quantum chips to near absolute zero, and how to correct errors when a disturbance inevitably erases the information in their qubits.
Other companies are taking a different approach. Microsoft’s chip uses topological qubits, which in theory are easier to manage. The technology remains unproven, however, and the company’s claims have generated criticism among researchers. Microsoft is confident it’s on the right path, however, and is continuing to release new papers. Darpa will evaluate its systems, too.
“We are confident in our approach to build a utility-scale quantum computer, which we believe we will achieve in years, not decades,” says Jason Zander, a Microsoft executive vice president responsible for the company’s quantum computing efforts. His team’s approach will lead to a usable 1-million-qubit computer that fits into a closet, he adds.
These diverging approaches are a fundamental concern for quantum computing: While the silicon transistor, invented in 1954, rapidly became the building block of all microchips, there’s no such consensus for quantum.
“At the moment, we’re tracking eight physical strata that can do quantum computing,” says Gartner’s Horvath. Some of these, like IBM’s, require temperatures colder than deep space. Others work at room temperature, but are nascent. There’s even one that uses diamond as a substrate; it’s extremely hard to manufacture.
Unlike the AI revolution, which few saw coming, the quantum one has been predicted for so long that big tech companies are keen not to miss it. Tellingly, Nvidia boss Jensen Huang declared in January that quantum computing was a long way off, sending associated stocks tumbling, then changed his tune in March, in time to announce his company’s own quantum efforts.
IBM’s Krishna welcomes the competition. Seeing dozens of other companies pursue the same goal only validates his assertion that this revolution is just around the corner.
“If it’s just us,” he told me, “you’ll begin to question—‘Why should I believe you?’”
Write to Christopher Mims at christopher.mims@wsj.com