Investors are throwing money at quantum startups. Maybe they should be looking at a more venerable player that has a lot of practice building things.
Half a century ago, a factory in Poughkeepsie, New York, cranked out computer hardware. The profits from mainframes financed pampered employees, scientific research and a dividend that made International Business Machines the most valuable company on the planet.
Now, a diminished IBM gets most of its revenue from soft things: computer programs and business services. But it’s at work on a new kind of machine that could return Poughkeepsie to its glory days. This is where it will assemble quantum computers, the magical devices designed to tackle mathematical challenges that would overwhelm an ordinary computer.
If quantum delivers on its promises, engineers will use it to make giant strides in the design of drugs, vaccines, batteries and chemicals. Last year Boston Consulting Group predicted that come 2040, quantum hardware and software providers will be taking in $90 billion to $170 billion of annual revenue.
IBM has been part of this rapidly evolving technology since the turn of the century. Leading its effort: Jay Gambetta, a 46-year-old physicist from Australia who oversees 3,000 employees on six continents doing research. He will not stint quantum, since he has spent his entire career in that field.
Gambetta joined IBM’s Watson Research Center, 39 miles south of the Poughkeepsie factory, in 2011 after postdoc years at Yale and then on the faculty at the University of Waterloo. He says, “While I like teaching, really I wanted to build.”
There are a lot of ways to build a qubit, the information-storing element of a quantum computer, and any might lead to a winner in the race to construct a useful machine. Light’s photons are quantized, a discovery that got Albert Einstein a Nobel prize, and make up qubits in some experimental computers. Ions—charged atoms—can be the basis of a quantum system. Yet another angle involves electric currents flowing in tiny superconducting wires deposited onto slivers of silicon. Within three years of Gambetta’s arrival at Watson, he and his colleagues decided to bet on that third option, turning away from photonics, trapped ions and other avenues of research.
The superconducting approach involves a chip chilled to a seventieth of a degree above absolute zero, a necessity for the superconductor to work and the delicate dance of electrons to be protected from corrupting thermal noise. The chip’s operating elements, called transmons, are controlled by pulses of microwaves. Its marching orders come from a conventional computer sitting nearby.
It helps that the unimaginably low temperature can be had with devices bought off the shelf, that chip manufacturing is something IBM can do in-house and that the microwaves, which are much like the microwaves running a cellphone, are old hat to electrical engineers. “We don’t have to reinvent things,” Gambetta says. “We leveraged 50 years of radar and microwave technology to make beautiful, clean microwave notes that we play.”
Alongside IBM and a few other giant enterprises researching quantum computing are a bevy of startups making pronouncements about their breakthroughs and their futures. They all have a lot of work to do before anything of great commercial value emerges. That hasn’t stopped investors from flinging money at them.
How to Play It
By William Baldwin
Two ancient vendors of data-processing machinery converge in a laboratory in Kobe, Japan. There a quantum computer made by IBM is married to a quick-witted (537 quadrillion double-precision operations per second) classical computer made by Fujitsu. You can enjoy a wild ride on shares in one of the many quantum startups promising dazzling results to arrive sometime in the next decade or two. Or you can take a tamer path. Fujitsu (traded in Tokyo and here as an ADR) and IBM will get an important piece of the quantum and artificial intelligence future but in the meantime get steady profits from mainframes, software and consulting.
William Baldwin is Forbes’ Investment Strategies columnist.
A firm in Hoboken, New Jersey, is one of those hot prospects. This corporation started out in the business of selling inkjet cartridges. That didn’t work out, so it switched to beverage distribution, which also fizzled. It changed its name to Quantum Computing and is selling photonics products. On its website: “Our vision: Put quantum into the hands of a billion people.” It was recently trading at 9,500 times revenue.
There are visions and there are working machines. IBM has quantum computers plugged in at the Poughkeepsie factory, at the research lab and in Europe and Asia. Scientists at Moderna, the Cleveland Clinic, Oak Ridge National Laboratory and many other institutions are running test programs on these machines, aiming to be ready with algorithms when faster, larger and more fault-tolerant quantum computers arrive.
IBM has good company betting on transmons: Google is taking the same tack. Might some entirely different scheme win the day? Gambetta considers this unlikely but he thinks about it. He makes a point of hiring engineers away from competitors pursuing other technologies, precisely so they can find flaws in whatever IBM is doing.
Some of these rivals have announced impressive results in small-scale experimental settings. But the leap to larger machines involves ever more demanding precision in the fabrication of quantum elements and the circuits that control them. Asks Gambetta rhetorically: “Do you have a plan to scale that technology? Are you building the fab with a lot of packaging?”
Then there is the problem that qubits are prone to error. As computing programs get more complex and involve more qubits, errors could accumulate to the point that the readout would be meaningless. Researchers are working on different ways to undo mistakes along the way, such as by having redundant qubits that keep one another in line. But this adds yet more complexity and more opportunities for failure.
Google has declared that it has a dramatically improved system for error correction. IBM has broadcast its answer to this part of the puzzle in a scientific journal. Says Gambetta: “I think we have the most transparent road map for error correction at scale.”
Quantum computing relies on two peculiarities of nature that were uncovered a century ago. One is that at a fine-grained level, the location and properties of objects are neither here nor there. Rather, they have when measured a certain probability of being here and a certain probability of being there. God rolls dice.
The other counterintuitive phenomenon is that two distinct objects can be married to each other—“entangled,” in the physicist’s term—even when separated. Measuring one determines a measurement of the other. This fact bothered Einstein. He dismissively called it spukhafte Fernwirkung, spooky action from afar.
The entanglement can be, but does not have to be, at a subatomic distance. This year’s physics Nobel went to scientists who demonstrated that it can take place over a stretch visible to the eye. With this in mind, IBM’s engineers will be pushing the microwave envelope. They are scaling up quantum computers in modular fashion, eventually wiring together, across several feet, two or more cabinets of superchilled, superconducting chips, with the qubits in one entangled with the qubits in the next. Einstein would be appalled.
Computers of the usual sort move in a deterministic fashion, a 0 becoming a 1 under a precise set of rules. Quantum computers make fuzzy moves. Impulses sent to a qubit nudge it one way or the other. If the nudges (called “gates”) are cleverly arranged, and if they act simultaneously on entangled qubits, their effect is to gradually push each qubit’s probable value toward either 0 or 1 in such a way that the qubits spell out a probable solution to a problem. This is a messy process, and it involves many cycles of nudges, with the superconducting chip frequently going back to a conventional computer for advice on what to do next.
Consider the problem Vanguard faces as it keeps its $44 billion Tax-Exempt Bond ETF up to date. There are at least 63,000 bonds to choose from. It selects 9,800, trying to get a mix that delivers a good yield while incurring the least possible risk. There is a certain hazard in lending money to Chicago at the same time as lending to Illinois, since both are at risk of being wrecked by pension-hungry unions. Numerous constraints, such as keeping the average maturity within a certain range, make the assignment mathematically difficult.
There is no way, at present, to find the optimal solution. Vanguard does the best it can, letting a conventional computer come up with a tolerably good answer. The process takes a few minutes and involves several hundred trillion calculations.
Quantum computing offers the prospect of getting better answers. In a recent experiment, IBM and Vanguard teamed up to explore what would happen if you wanted to optimally select from 109 securities. To try out each possible combination is unrealistic; on Vanguard’s computer it would require a run time somewhat longer than the age of the universe.
The Vault
Industrial Revolutionary
A century before charging into quantum computing, International Business Machines was helping drive an earlier revolution, ushering businesses into the mid–20th century with a lineup of hot new tech like “scales, slicing machines, time recorders, tabulating and cost-keeping machines.”
Typewriters have become virtually a necessity even for the small business man, while cash registers and adding machines, once used only by large concerns, are gradually becoming almost as necessary as typewriters. Recent inventions have so broadened the field of calculating and book-keeping machines, that in large concerns few records are kept by manual labor. Machines have been made so simple and relatively so inexpensive, they steadily are replacing old-fashioned bookkeepers. . . .
[IBM] is a holding concern, its three principal subsidiaries being the Dayton Scale Company, the Tabulating Machine Company and the International Time Recording Company. Each is recognized as a pioneer in its special division, so that in the fields it serves, the parent company is regarded in the trade as one of the foremost.
—Forbes, October 1, 1928
The quantum machine did not have to look at each combination sequentially. It was, in effect, looking at all possibilities at once, humming its way on those microwave notes toward the optimum. It groped toward an answer after 4,200 gate operations.
IBM has to deliver a lot more to be useful to a client like Vanguard. In its vision: in 2029, a room-sized fault-tolerant modular computer in Poughkeepsie that can run 100 million gates. Sometime before then, Gambetta says, smaller machines, working alongside classical computers, will be beating out purely classical computers at practical tasks like portfolio optimization. IBM already has in hand $1 billion of commitments for quantum services.
It would be easy to picture IBM as a stodgy firm good at delivering reliability to banks and airlines but unlikely to beat out nimble startups at the cutting edge. But note this: After a century of being run mostly by salesmen, the company now has as chief executive Arvind Krishna, a Ph.D. electrical engineer who used to have Gambetta’s job.
Could that firm in Hoboken, with its photons, race ahead of IBM and its transmons? Could happen. Or maybe it will go back to distributing sodas.
More from Forbes
Source: https://www.forbes.com/sites/baldwin/2025/11/25/inside-ibms-quest-to-win-the-quantum-computer-race/