Quantum Can’t Do Much yet but Banks Can’t Afford to Wait

It’s overhyped because the commercial applications most people imagine—the turbocharged artificial intelligence (AI)—and magical optimization, are largely fiction.

It’s underestimated because the one thing quantum machines will almost certainly be able to do is break the cryptographic locks that secure the modern financial system. And migrating those locks will take years.

“The time to start thinking about migrating to quantum-resistant methods of encryption is now,” said Professor Scott Aaronson, who recently joined StarkWare as scientific advisor, during a conversation hosted by PYMNTS CEO Karen Webster.

Not next year. Not when a working machine appears. Now, he said, because the transition itself is the bottleneck.

That’s the paradox at the center of quantum computing in 2026 as Aaronson sees it. The technology is nowhere near delivering on its grandest promises. But its most dangerous capability, codebreaking, doesn’t need to arrive tomorrow to demand action today.

The Threat That Doesn’t Need to Be Imminent to Be Urgent

Consider the timeline problem. Aaronson said that even optimistic estimates place practical quantum attacks five to ten years out. But implementing a cryptographic transition across a major financial institution, one that touches every protocol, every system, every vendor relationship, could easily consume that same window.

The margin for error is essentially zero.

Worse, the attack surface is democratized. Unlike, say, a nation-state’s nuclear program, quantum computers will likely be accessed through the cloud. That changes the risk calculus entirely.

“At some point they will be useful for attacking cryptography,” Aaronson said. “And at that point, no one quite knows how to monitor all the requests coming in to see which ones are trying to sneak in breaking a cryptographic code.”

This isn’t a theoretical exercise in contingency planning. It’s a migration problem with a hard, if fuzzy, deadline. And the work is enormous.

A Computational Lemon—and a Very Small Glass of Lemonade

So if the threat is real, what about the opportunity? Here’s where Aaronson said the story can get uncomfortable for anyone selling quantum’s future.

The honest scientific consensus is that quantum computing’s credible commercial sweet spot is remarkably narrow.

Simulating quantum systems. Drug discovery. Battery chemistry. High-temperature superconductors. Industrial chemical reactions.

These are problems where classical computers genuinely struggle, because the size of the quantum state grows exponentially with each interacting particle.

“Nature has given us this computational lemon,” Aaronson said. “Why don’t we make lemonade out of it?”

Today’s machines, roughly 100 qubits running a few thousand operations, he says, are beginning to simulate simplified models. In narrow cases, they may even edge past classical methods. But “edge past” is the operative phrase. Transformative applications will require far larger, fault-tolerant systems with robust error correction, and the overhead to run such a system is staggering. A modest quantum advantage doesn’t easily justify building one.

That hasn’t stopped the pitch deck industrial complex, Aaronson said. Startups continue to hawk quantum-enhanced everything, from handwriting recognition to portfolio optimization to neural network training, with evangelistic fervor.

Aaronson’s assessment was blunt: “People eat that up with mustard. But to the scientists, most of this is worthless. If you’re not actually beating a classical computer, then it’s not interesting to us.”

The AI Question: Asymmetry in Both Directions

An inevitable question arises: What happens when quantum meets AI?

The relationship turns out to be lopsided, but not in the direction most people assume.

AI is already accelerating quantum research. Neural networks are improving error correction decoding, and deep learning is optimizing circuit design.

AI is helping quantum get better.

Is quantum returning the favor to AI? Much less likely, Aaronson posited. The reason is structural.

Quantum algorithms work by exploiting probability amplitudes, the complex numbers that can interfere constructively or destructively, to cancel out wrong answers and amplify the right one. It’s an intricate choreography that only works for problems with very specific mathematical structures.

Aaronson likened it to “this bizarre hammer that you have to find some nail that it can hit.”

Claims that quantum computers will revolutionize machine learning don’t hold up to his scrutiny or those of his fellow scientists.

“I don’t think the science supports it,” Aaronson said, adding that theoretical advantages like Grover’s algorithm exist on paper, but they’re modest and tend to get swallowed by the overhead of error correction.

He said he sees an even more fundamental problem: Classical computing won’t stand still while quantum catches up.

“Classical computing is a moving target,” Aaronson said. Again and again, claimed quantum advantages have evaporated once classical researchers sharpened their own algorithms. “You have to beat it. That’s the hard thing in this field.”

Compare that to AI, whose explosion comes from sheer breadth of application. “There seems to be no limit to what it is good for,” Aaronson said. “Any intellectual work that any human does … It has become harder and harder to find examples where AI could not do the same thing.”

Narrow Doesn’t Mean Trivial

Quantum computing’s impact, even in the best case, will be narrow. But Aaronson stressed that narrow is not the same as trivial. A handful of genuine breakthroughs in drug design or materials science could catalyze billion-dollar industries. The hammer is strange, but the right nails are worth hitting.

For financial institutions, though, the calculus is different. They don’t need to bet on quantum’s upside. They need to defend against its downside. And the window to do that is closing faster than the technology itself is advancing. The commercial payoff may be a decade away. The security risk isn’t.

That’s the brutal math Aaronson. And solving for it, he said, starts now.