When will quantum computers be useful?
The short answer
Nobody knows for certain. Today's quantum computers are noisy and mostly used for research, and they have only beaten ordinary computers on carefully chosen test problems. Most experts expect the first genuinely useful results in chemistry and materials simulation, with large error-corrected machines able to run algorithms like Shor's further off. Several companies aim for fault-tolerant machines around the end of this decade, but treat any precise date with caution.
Learn this step by step in the Future of Quantum Computing course, in the free Q-Learn app. Get it on Google PlayGet Q-Learn
Where we are now
Today's machines belong to what the physicist John Preskill named the NISQ era, for Noisy Intermediate-Scale Quantum. They have tens to a few thousand physical qubits, errors on every operation, and limited error correction. They are excellent tools for research, and anyone can try one through cloud services from IBM, Amazon, Microsoft and others.
There have been striking demonstrations. In 2019 Google's Sycamore chip performed a sampling task it estimated would take a supercomputer thousands of years, and in 2024 its Willow chip showed error correction improving as it was scaled up. But the 2019 task had no practical use, and better classical algorithms later narrowed the gap considerably.
What "useful" really means
The milestone that matters is practical quantum advantage: solving a problem that someone actually cares about, faster, cheaper or more accurately than the best classical method. Not a contrived benchmark, and not a comparison against a slow classical method. Claims of this kind have started to appear for narrow scientific problems, but none is yet widely accepted as a clear, general win.
What is likely to come first
- Chemistry and materials science. Simulating molecules is what quantum computers are naturally suited to, and it is where most experts expect the first real advantage: catalysts, batteries, new materials and drug discovery.
- Physics research. Simulating other quantum systems is already producing scientifically interesting results.
- Optimisation and finance. Heavily researched, but a clear advantage over the best classical methods has not been shown.
- Machine learning. Promising in theory, but loading large amounts of classical data into a quantum computer is a major bottleneck.
- Breaking encryption. Needs a large fault-tolerant machine, so it is further away than the applications above. See Can quantum computers break encryption?
What has to happen
- Better qubits: lower error rates and longer coherence times.
- Error correction at scale: many reliable logical qubits, not just a handful.
- More qubits, well connected: probably through modular machines linked together.
- More algorithms: proven speedups exist for only a limited set of problems, and finding new ones is an active field.
What the roadmaps say
Several companies have published roadmaps. IBM, for example, aims to build a large-scale fault-tolerant quantum computer by 2029, and others including Google and Quantinuum have set similar goals for the end of the decade. Roadmaps are targets, not guarantees, and the history of the field includes plenty of missed dates. They are also a sign of how seriously the industry now takes the engineering challenge.
How to spot hype
When you read a quantum computing headline, ask:
- Is the result compared against the best classical method, or just a convenient one?
- Are the qubits physical or logical? A thousand noisy physical qubits is very different from a thousand error-corrected ones.
- Is the problem useful, or a benchmark designed to suit quantum hardware?
- Has the work been peer reviewed or independently checked?
Why learn about it now
Quantum computing is moving from physics labs into engineering, and the people who understand the basics, what qubits, superposition and entanglement really are, will be best placed to judge the claims and spot the opportunities as they arrive.