Separating the plausible from the pitch deck
Almost every industry has been told quantum computing will transform it. Some of those claims rest on a real algorithmic argument; most rest on the observation that the industry has hard optimisation problems, which is not the same thing. This hub covers the applications people are actually working on, and is explicit about which stage each one has reached.
Where the case is strongest
- Chemistry and materials — the best-founded application, because the problem is natively quantum. Catalyst design, battery chemistry, nitrogen fixation. Still waiting on error-corrected hardware, but the argument does not depend on a contested speedup.
- Quantum sensing and metrology — commercially deployed today, and consistently underreported. Quantum gravimeters, magnetometers and atomic clocks are working instruments used in navigation, medical imaging and mineral survey. They are not quantum computers, which is why they rarely make the same headlines.
- Cryptography — the one application with a firm deadline attached, covered in Quantum Cryptography & Security.
Where the case is contested
- Finance — portfolio optimisation, derivative pricing, risk simulation. Quantum Monte Carlo methods offer a quadratic speedup in theory; whether that survives the overhead of error correction on real instrument sizes is unresolved.
- Logistics and optimisation — routing, scheduling, supply chains. The most heavily marketed application and the weakest theoretically: classical solvers are extremely good, and no general quantum speedup for practical optimisation has been demonstrated.
- Drug discovery — real in the chemistry-simulation sense, oversold when presented as end-to-end drug design.
- Machine learning — see Quantum Machine Learning for why the data-loading problem keeps undercutting the claims.
The quantum internet
A separate track from computation: networking quantum devices with entanglement, which needs quantum repeaters that do not yet exist at scale. Applications include distributed computing, blind quantum computation and networked sensors. Early multi-node networks are running in labs and over metropolitan fibre.
A question worth asking
For any claimed application, ask what the best classical method achieves today, how large an error-corrected machine the quantum version needs, and whether the comparison came from the vendor. Those three answers explain most of the distance between a press release and a result.
