What NISQ actually means, and why every quantum computer fits into that category?

[Tracey99]

The Quantum Insider published a solid explainer this week on NISQ, or Noisy Intermediate Scale Quantum, the term physicist John Preskill coined back in 2018 to describe the quantum computers actually being built rather than the idealized, fully fault tolerant machines that dominated earlier theoretical discussions. The piece makes a point worth sitting with, every commercially available quantum computer today, IBM's superconducting systems, Google's processors, IonQ's trapped ion hardware, all of it falls into this same NISQ category regardless of how a given press release chooses to frame it.

Each letter in the acronym does real work. Noisy refers to the fact that qubits are extremely sensitive to their environment, with stray electromagnetic fields, temperature fluctuations, and imperfect control signals all introducing errors that accumulate as a computation runs longer. Intermediate scale describes systems in the tens to low thousands of qubits, large enough to be genuinely hard to simulate on a classical computer but far smaller than the millions of physical qubits full fault tolerance is expected to eventually require. Quantum just confirms these machines actually exploit superposition, entanglement, and interference rather than being classical computers dressed up with quantum branding.

The practical limitations follow directly from that noise and scale. Circuit depth is limited because errors pile up with every additional gate applied, gate error rates on current systems typically run somewhere between 0.1 and 1 percent per operation, and full error correction remains out of reach since it requires far more high quality qubits than any current system has. Techniques like zero noise extrapolation and probabilistic error cancellation can meaningfully stretch how much useful computation NISQ hardware can squeeze out before noise dominates, extending effective circuit depth from hundreds of operations into the thousands or occasionally tens of thousands.

The article also makes clear the NISQ era, while still ongoing, is starting to show real cracks at the hardware level. Google's Willow chip demonstrated below threshold error correction back in December 2024, meaning adding more physical qubits to an error correction code actually reduced the logical error rate for the first time on real hardware. Quantinuum followed with a fully fault tolerant gate set in mid 2025 and 94 error protected logical qubits by March 2026. Industry roadmaps are broadly converging on the late 2020s for genuine fault tolerant performance, though the piece is careful to frame this as a gradual transition rather than a single clean cutover point

What Is NISQ Quantum Computing: The Current Era of Quantum Machinesthequantuminsider.com

[Carol15]

Error mitigation versus error correction is a distinction that gets flattened together constantly in casual reporting, and this piece does a decent job keeping them separate. Mitigation is a statistical, after the fact trick for inferring what a noiseless result probably would have looked like, it doesn't actually fix anything happening during the computation itself.

Correction is the real deal, actively detecting and fixing errors as they occur using redundant encoding across many physical qubits. Mitigation buys useful time and extends what NISQ hardware can accomplish in the meantime, but it's fundamentally a workaround rather than a solution, and conflating the two makes it harder to judge how close the field actually is to genuine fault tolerance

[BankHolidayBlues87]

0.1 to 1 percent gate error rates sound tiny until you remember how quickly those errors compound across even a moderately deep circuit. A few hundred sequential gates at that error rate and you're looking at a computation that's basically pure noise by the end

[Caitlin89]

The Willow below threshold result genuinely deserves more attention as a milestone than it usually gets outside specialist circles. Confirming that adding more physical qubits to an error correction code actually reduces the logical error rate, rather than just adding more places for errors to occur, is the specific inflection point the whole field has been waiting on for years. Before that result, it was theoretically obvious error correction should eventually work this way, but nobody had confirmed it experimentally at meaningful scale. That's a properly different kind of validation than another qubit count record

[Hollow Ronan]

What strikes me is how quietly the industry consensus target date has converged around 2028 to 2029 across completely different hardware approaches, IBM, IQM, Oxford Quantum Circuits, and Quantinuum all landing in roughly the same window despite pursuing very different physical qubit implementations. That kind of convergence across competing approaches with different underlying physics is either a properly good sign that the engineering challenges are becoming well understood industry wide, or it's just everyone anchoring on the same optimistic public messaging without much independent basis for the specific date. Hard to tell which from the outside without seeing the actual internal engineering data behind each roadmap. Would love to see an independent audit of how realistic each company's specific timeline actually is rather than just taking the announced dates at face value

[Violet Tiger]

Nice that the piece includes an actual comparison table between NISQ and fault tolerant systems rather than just describing each in prose. Those side by side qubit count, error rate, and application comparisons make the practical gap between where we are now and where the field wants to go a lot more concrete. More explainers should include a table like that instead of burying the comparison in paragraphs

[CaptainStatic56]

The quantum utility framing IBM introduced with their 127 qubit Eagle processor result is worth unpacking a bit more than this piece has room for. Producing results for a physics simulation that classical brute force methods couldn't reliably verify is a meaningfully different claim than quantum advantage or quantum supremacy, since it doesn't require proving no classical method could ever match the result, just that brute force verification specifically becomes impractical at that scale.

That's a more modest and arguably more honest claim than some of the flashier advantage announcements that get more headlines. Google's more recent demonstration with the Willow processor follows a similar pattern of narrower, more carefully scoped claims rather than sweeping supremacy declarations. It suggests the field has learned something from earlier disputes over what actually counts as a legitimate advantage claim. Modest, well scoped claims that hold up under scrutiny are probably better for the field's credibility long term than bold claims that later get walked back

[Hollow Panther]

Good reminder that every splashy quantum computing headline right now is still fundamentally describing a NISQ machine underneath the marketing. Useful filter to apply whenever a press release claims some dramatic breakthrough without mentioning error correction status specifically

[GradientHydra]

The variational algorithms section undersells just how much active research still goes into finding problems these hybrid approaches can actually beat classical methods on. Demonstrating genuine advantage with VQE or QAOA specifically remains a real open question rather than settled science