Banks are betting that quantum computing can find them an investment edge

From his office in Zurich, Stefan Woerner can make a few keyboard strokes that may have the potential to transform how the financial sector processes data.

Woerner, who leads IBM's quantum finance efforts, regularly logs on from Zurich to one of 15 machines maintained in Poughkeepsie, New York, at IBM's Quantum Computation Center. He uses these quantum computers to run small financial calculations that estimate a stock portfolio's risk, for example, or forecast the price of an asset.

Actually, an ordinary desktop computer could easily handle these calculations; they're based on market models far too simplistic to be useful for a bank. But Woerner's team, which has partnered with banks such as JPMorgan Chase and Barclays, runs these small demonstrations to explore how quantum computing might serve financial institutions. Executives want to know whether a decade from now, the emerging technology could help them reap bigger profits or minimize losses.

"This is a very interesting journey," Woerner says, "but we are at the very beginning."

Right now, quantum computing's advantages over conventional computers are still almost entirely theoretical. But that hasn't stymied financiers' curiosity. Just this year, IBM announced new partnerships with Goldman Sachs and Wells Fargo. Goldman, along with Citigroup, are also lead investors of quantum software startup QC Ware, based in Palo Alto. The collaborations have extended across the globe, with IBM also working with Mizuho and Mitsubishi Financial Group in Japan, and quantum startup Multiverse Computing partnering with BBVA in Spain.

Some experts think that quantum computers have the potential to speed up financial computations by more than a thousand times, says Ning Shen, managing director of quantitative research at JPMorgan Chase.

They have mathematical reasons for suspecting that quantum computers will be useful for finance, and physicist Roman Orus explains it like this: Change the price of oil or corn or silicon, for example, and other commodities' prices shift, too, sometimes in unexpected ways because of interdependencies in a complex global supply chain. Banks would like to be able to accurately predict these shifts, but "these are problems very difficult to be solved with a classical computer," says Orus, who founded Multiverse Computing last year.

Quantum computers, on the other hand, should be able to simulate markets much more accurately by virtue of their mathematical architecture. The math of quantum mechanics naturally describes the behavior of intricately correlated objects, such as the relationships among electrons pushing and pulling on each other in a material. The tangle of industries and economies generate ups and downs in a financial market not unlike how an intricate network of electrons in a material produces superconductivity, Orus says. According to Orus, simulating a financial market accurately is comparably difficult to modeling complex molecules, and both are among the most difficult problems in computer science.

In particular, financial institutions are investigating the machines' potential to improve machine learning, optimization and so-called Monte Carlo simulations, which are used to forecast future markets, Shen says. Banks perform these kinds of calculations on a daily basis.

To test these ideas, researchers run experiments. Shen, for example, has worked with Woerner at IBM to predict the price of options and portfolios of options using a quantum computer. They made these predictions by performing Monte Carlo simulations, a technique for forecasting investment values that involves simulating millions of scenarios: changes in interest rates or taxes, for example. Then, they make a forecast by taking a statistical average of these simulated futures.

Woerner, collaborating with Barclays, has also worked on a simple algorithm for optimizing transaction settlement, a process in which banks trade stocks mediated by a third-party clearing house. Clearing houses settle more than a quadrillion dollars in these transactions per year, and the financial industry wants to speed up the process, Woerner says. Meanwhile, quantum machine learning researchers are developing algorithms for recognizing patterns in generic data, which could be used to help predict markets, Orus says.

It's unclear which finance algorithms will deliver first, or even if they will at all. The current work is largely exploratory, and neither Goldman Sachs nor JPMorgan Chase expects to employ quantum computers in real-time bank operations in the next few years. Goldman Sachs considers quantum computing as a "longer-term" project, says Jeremy Glick, the bank's global head of R&D engineering. Right now, one of their goals is simply to foster communication between banks and quantum experts. "These are research collaborations," says Matt Johnson, the CEO of QC Ware.

In a way, the industry's speculative research into quantum computing is just business as usual. Big banks often explore emerging technologies in pursuit of an advantage over their competitors, says mathematician Kees Oosterlee of Centrum Wiskunde & Informatica in the Netherlands. About a decade ago, they experimented with graphics processing units, which they now rely on regularly to perform Monte Carlo simulations. "They start with small experiments, and only when it gets more promising and more realistic, they will invest more in that technology," Oosterlee says.

But it's no guarantee that quantum computing will pay off like GPUs did, and researchers have set their ambitions cautiously. "The goal is to understand what kind of problems the quantum computer can solve," Shen says.