Over the past decade, quantum computing has grown into a billion-dollar industry. Everyone seems to be investing in it, from tech giants, such as IBM and Google, to the US military.

But Ignacio Cirac at the Max Planck Institute of Quantum Optics in Germany, a pioneer of the technology, has a more sober assessment. “A quantum computer is something that at the moment does not exist,” he says. That is because building one that actually works – and is practical to use – is incredibly difficult.

Rather than the “bits” of conventional machines, these computers use quantum bits, or qubits, to encode information. These can be made in several ways, from tiny superconducting circuits to extremely cold atoms, but all of them are complex to build.

The upside is that their quantum properties can be used to do certain kinds of computation more quickly than standard computers.

Such speed-ups are attractive for a range of problems that normal computers struggle with, from simulating exotic physics systems to efficiently scheduling passenger flights or grocery deliveries to supermarkets. Five years ago, it seemed quantum computers would ameliorate these and many other computational challenges.

Today, the situation is a lot more nuanced. Progress in building ever bigger quantum computers has, admittedly, been stunning, with several companies developing machines with more than 1000 qubits. But this has also revealed impossible-to-ignore difficulties.

One major problem is that, as these computers get larger, they tend to make more errors, and finding ways to prevent or fix these has proven to be harder than expected. Last year, Google’s researchers made the most notable dent in this problem so far, but even so, fully fledged, useful quantum computers aren’t here yet – as Cirac points out.

Because of this, the list of realistic applications for these machines may be shorter than we once hoped. Weigh the cost of building one against the smaller-than-imagined savings it could deliver, and, for many use cases, it may not make economic sense. “The biggest misconception is that a quantum computer can accelerate any problem,” says Cirac.

So, which problems might still benefit from quantum computation? Quantum computers could break the cryptography systems we currently use for secure communication, and this makes the technology interesting to governments and other institutions whose security could be imperiled by it, says Scott Aaronson at the University of Texas at Austin.

Another place where quantum computers should still be useful is in modelling materials and chemical reactions. This is because quantum computers, themselves a system of quantum objects, are perfectly suited to simulate other quantum systems, such as electrons, atoms and molecules.

“These will be simplified models; they won’t represent real materials. But if you design the system appropriately, they’ll have enough properties of the real materials that you can learn something about their physics,” says Daniel Gottesman at the University of Maryland.

Quantum chemistry simulations may sound more niche than scheduling flights, but some of the possible outcomes – finding a room-temperature superconductor, say – would be transformative.

The extent to which all this can truly be realised is significantly dependent on quantum algorithms, the instructions that tell quantum computers how to run – and help correct those pesky errors. This is a challenging new field that Vedran Dunjko at Leiden University in the Netherlands says is forcing researchers like him to confront fundamental questions about what information and computing are.

“This provides an amazing motivation to study the hardness of problems and the power of computing devices,” says Dunjko. “For me, this would be reason enough to dedicate a significant fraction of my life to these questions.”