Why banks need to start planning to use quantum computing

Quantum computers will eventually present the threat of breaking the encryption algorithms that banks and credit unions use today. But that is a mitigatable threat, and banks have plenty of uses for quantum computers that have driven the world's largest institutions to invest in the technology.

These quantum computing uses are purely hypothetical at the moment, according to one expert, who said no quantum computer today can solve any real-world problems faster than a classical computer can. But as testing on the technology continues, small- and medium-size banks can expect a clearer list of the potential ways they can use quantum computing in the future.

So far, that list of potential applications includes risk analysis, investment portfolio construction and financial crime monitoring, and it may grow over time.

In recent years, J.P. Morgan Chase, Bank of America, Barclays and Morgan Stanley have all built teams to investigate quantum computing technology and its applications. Ally announced last year it was testing use of the technology to bring efficiency to investment portfolio construction. Wells Fargo announced a partnership with IBM and MIT in 2019 to investigate quantum computing technology, and HSBC made a similar announcement the next year.

"We need to embrace this technology, in line with the bank's innovation agenda, by keeping up with the latest developments and growing our internal knowledge to increase our readiness for the post-quantum world," said Gustavo Ordonez-Sanz, head of economic capital analytics and global risk innovation lead for HSBC.

Many observers track the progress of quantum computers by how quickly their limitations are lifted. For example, the number of qubits — the quantum equivalent of the bits used by classical computers — that a quantum computer can manipulate has risen over time.

The current record for qubits in a quantum computer stands at 433, which IBM achieved last year. That broke the 127-qubit record set by IBM the year prior. However, IBM's quantum computers (and all others with a large number of qubits so far) have an error rate that makes them ineffective for many uses.

Observers also track this error rate to gauge when quantum computers might reliably outperform classical computers on various tasks. According to a February blog post by Google's Quantum AI team, error-corrected quantum computers will require error rates of about one in a million, but today's quantum computers typically have an error rate around one in a thousand.

Once quantum computing hardware has reached an adequate level of error correction, qubits and other metrics, banks will have a new worry to address: How to actually program a quantum computer.

Quantum computer software used in labs today differs greatly from the programming currently taught in schools and universities. While companies like Google and IBM use familiar programming languages like Python to enable people to experiment with their quantum computers, these programming languages act merely as a wrapper around a much different set of instructions that would be alien to any programmer who has not specifically studied quantum computers.

Some companies offer services that help bridge the knowledge gap that programmers have between classical programming and quantum programming. One such company is the Singapore-based Horizon Computing, founded and led by Joe Fitzsimons.

The lack of quantum computing expertise and education today could present a challenge as the technology develops, Fitzsimons said, and that will present smaller banks with a unique challenge.

"Places like J.P. Morgan Chase and Goldman Sachs have quantum teams," Fitzsimons said, and those teams consist of "experienced people that have been in the quantum computing community for a significant amount of time." However, for smaller institutions, building up those teams is "not necessarily an option" because "there's definitely going to be — and currently is — a significant talent bottleneck. There's just not enough people with experience."

Banks will not need to replace all their classical computers with quantum computers, though. They will not be able to speed up every computational task, Fitzsimons said. But for a set of computational tasks, the speed-up quantum computers offer will be significant, and in some cases, exponential.

Quantum computers will enable banks to improve their loan underwriting models, more accurately calculate loan prepayment and default risks and process more data inputs in their marketing models, according to a blog post published this month by the American Bankers Association. Quantum computing could also help banks improve their risk profiling, optimize their trading strategies and increase their ability to detect fraud and other financial crimes, according to IBM.

In general, problems that involve mathematical optimizations and repeated random sampling are the most viable for quantum speed-ups, according to Fitzsimons, but he expects quantum computers that can take in large amounts of data to take a long time to develop.

"Figuring out how you're going to take advantage of quantum computing — that's an important problem, and it's not something you can solve in a rush," Fitzsimons said.