Quantum Computers Today: The NISQ Era
If qubits are so powerful, why do today's quantum computers have only a few dozen or a few hundred of them — while your laptop has billions of transistors? The short answer: a qubit is a prima donna. It stays quantum only while it is protected from the world, and protecting one qubit is hard; protecting a thousand, each wired up and talking to its neighbors, is enormously harder.
Three forces fight you. First, control: every qubit needs its own control lines, and the wiring and calibration complexity grows with each qubit you add. Second, noise: heat, vibration, stray electromagnetic fields — any disturbance leaks information out of the qubit and destroys its superposition. Physicists call this decoherence, and it is the central enemy of the field. Third, cold: many qubit designs (superconducting circuits, for example) only behave quantum-mechanically near absolute zero, so the whole processor lives inside a dilution refrigerator colder than deep space.
That is why we chase "only a few qubits" — and why the era we live in has a name: NISQ, Noisy Intermediate-Scale Quantum (a term coined by John Preskill in 2018). Noisy: error rates are high and error-correcting codes are not yet in practical use — every operation is a little bit wrong. Intermediate-scale: qubit counts are in the tens to hundreds — far more than a few, far fewer than the millions that fault-tolerant algorithms will need.
The race is crowded and well funded: Google, IBM, Intel, Microsoft, D-Wave, Rigetti, IonQ and many others are building machines on different qubit technologies, and investment has surged as governments and companies bet on the field. But the number that matters most is not the headline qubit count — it is how many of those qubits are good qubits.
A hundred noisy qubits can lose to ten excellent ones. That is why the field invented better yardsticks than raw count: quantum volume folds qubit count and error rates into one number, and CLOPS — circuit layer operations per second — measures how fast a machine actually executes circuits. The right question about any quantum computer is never "how many qubits?" but "what useful circuit can it run before the noise takes over?"
(See also: Decoherence, Quantum computing in 10 questions.)
Go deeper — the math & the rigor
Quantum volume (QV) tries to answer "how big a successful computation can this machine run?" in a single number. The definition: the machine can reliably run random square circuits up to some width and depth \(d\), and the quantum volume is \[QV = 2^{\min(N,\,d(N))}\] where \(N\) is the number of qubits and \(d(N)\) is the largest circuit depth the machine achieves with acceptable error. The \(\min\) is the point: width without depth is useless — a thousand qubits that decohere after two layers cannot run anything interesting.
Depth matters because every gate is a chance to fail. If each two-qubit gate fails with probability \(p\), a circuit with many gates succeeds with probability roughly \((1-p)^{\text{number of gates}}\) — exponential decay in the number of operations. Improving the error rate from 1% to 0.1% does not just make circuits slightly better; it multiplies the achievable depth roughly tenfold, and because QV is exponential in depth, the quantum volume can jump by orders of magnitude. This is why a modest hardware improvement can look like a huge leap in QV — and why vendors chase fidelity as hard as qubit count.
CLOPS — circuit layer operations per second — measures something QV does not: speed. A machine might have an impressive quantum volume but take an hour to run one circuit; CLOPS counts how many circuit layers it executes per second, folding in gate times, measurement speed, and the classical overhead of the control stack. It is the quantum analogue of asking not just how powerful a processor is, but how fast it actually runs your program.
Together the metrics tell the NISQ story honestly: we have machines with over a thousand physical qubits announced, yet quantum volumes that correspond to a few dozen effective qubits — and no error correction in production yet. The NISQ era is the era of making every qubit count: shallow circuits, clever compilation, and error mitigation, until fault tolerance arrives.
Key takeaways
- A qubit stays quantum only while isolated; control complexity, noise (decoherence), and cooling make scaling hard.
- NISQ = Noisy Intermediate-Scale Quantum: high error rates with no error correction in use, and qubit counts in the tens to hundreds.
- Headline qubit counts mislead — what matters is how many qubits are good qubits.
- Quantum volume, QV = 2^min(N, d(N)), combines qubit count with achievable circuit depth; depth is limited by error rates.
- CLOPS (circuit layer operations per second) measures how fast a machine actually executes circuits.
Check your understanding
Q1.What does 'Noisy' mean in NISQ?
NISQ devices run without practical error correction, so every operation carries a real chance of failure.
Q2.In QV = 2^min(N, d(N)), why does the min matter?
A huge qubit count means nothing if noise kills the computation before any depth is achieved.
Q3.What does CLOPS measure?
CLOPS (circuit layer operations per second) folds gate times, measurement, and control overhead into a speed metric.
References
The books, papers, and articles this lesson drew on — with a note on what each one was used for.
Tip: press ← / → to turn pages.