Quantum computing occupies an unusual position: the underlying research is genuinely advancing, the engineering milestones are real, and the practical applications remain almost entirely prospective.
What has been achieved
Machines with meaningful numbers of physical qubits, operating stably enough to run algorithms. Demonstrations of error correction, where multiple physical qubits combine into a logical one that survives longer than any of its parts.
That second point is the important one. Error correction working at all was the fundamental question, and answering it moved the field from theoretical to engineering.
What has not
Any commercially valuable computation performed faster or cheaper than a classical computer could do it. Not one, despite considerable effort to find a case.
The demonstrations of quantum advantage have been on problems constructed specifically to suit quantum hardware, with no application outside the demonstration. That is legitimate science and it is not a product.
The error correction overhead
Current estimates require somewhere in the range of a thousand physical qubits per logical qubit for a useful error rate. An algorithm needing a few thousand logical qubits therefore needs millions of physical ones.
The largest machines today have thousands. The gap is not a matter of iteration; it requires the physical qubit count to grow by roughly three orders of magnitude.
The cryptography question
A sufficiently large quantum computer would break the public-key cryptography protecting most internet traffic. No such machine exists and none is close.
The response is nonetheless urgent and correct. Encrypted data captured today can be stored and decrypted later, so anything requiring long-term confidentiality needs post-quantum algorithms now. Standards exist and migration has begun.
A reasonable expectation
Useful quantum computation, if it arrives, arrives as a specialised service for a small set of problems – molecular simulation, certain optimisation classes – accessed remotely. Not as general-purpose computing, and not on any timeline anyone can currently give with confidence.