What Today's Machines Actually Do

Real quantum computers exist and you can rent one in the cloud — but they’re noisy, uncorrected, and run only shallow circuits: the Wright Flyer, not the 747.

Quantum Computing · Lesson 16 · 10 min read

Headlines swing wildly: one week “quantum computers achieve supremacy!”, the next “quantum computing is a scam that does nothing.” Both miss the truth. Real quantum computers exist today — you can even rent time on one over the cloud — but what can they actually do right now, versus what’s still years away? Hold the question. Separating those two is the single most useful skill you can carry out of this whole track.

The NISQ era: noisy, intermediate-scale, not corrected

Today’s machines are NISQ — Noisy Intermediate-Scale Quantum. They have enough qubits to be genuinely interesting, but too few and far too noisy to run full error correction at scale. With no error correction (L15) and short coherence times (L14), they can only run shallow circuits — a handful of gate layers before noise swamps the answer. That single fact — shallow only — shapes everything they can and can’t do.

What they can and can’t do

Can: prototype small algorithms, run small physics and chemistry experiments, and perform quantum advantage demonstrations — doing some contrived task (often a sampling problem) faster than the best classical computer. Those demos prove the hardware is truly quantum, even when the task itself is useless. Can’t (yet): break real encryption, simulate industrially relevant molecules, or run any deep algorithm — all of which need vastly more gates, and therefore error correction and millions of qubits, than NISQ machines have.

Where it’s really headed (optimistic and honest)

The milestone everyone chases is fault tolerance: the first logical qubits that outlive their physical parts, error rates dropping below threshold, then scaling up. Progress is real and fast — this is the early-transistor stage, not science fiction and not vaporware. The world-changing uses (chemistry first, then code-breaking) arrive with scaled, error-corrected machines, not with today’s NISQ chips. Holding both truths at once — real now, transformative later — is what inoculates you against hype and doom.

Sorting four claims about today’s machines:
1. “A quantum computer did in minutes a sampling task a classical supercomputer would take millennia to match.” → Plausibly true (an advantage demo) — BUT the task is contrived; it proves the hardware is quantum, not that it’s useful.
2. “You can run a real quantum algorithm on a cloud quantum computer today.” → True — but only shallow ones; deep circuits drown in noise.
3. “A quantum computer just factored a real RSA key.” → False today — that needs millions of error-corrected qubits.
4. “Quantum computing is useless hype.” → False — real machines hit genuine milestones (first logical qubits, falling error rates).
5. The skill: for any headline ask, “is this a shallow NISQ demo / contrived task, or a claim of deep, useful, error-corrected computation?” That one question sorts truth from hype in seconds.