Every quantum computer built so far is a noisy machine. Errors accumulate at rates many orders of magnitude above what useful computation requires, and no amount of physical engineering has closed that gap. The only known path is quantum error correction: encode one logical qubit across many physical qubits, detect errors without measuring the data, and correct them faster than they occur.
The pivotal concept is the threshold theorem: if physical error rates sit below a critical value, adding more hardware makes the logical qubit better instead of worse. For twenty years the field chased that threshold. It has now been demonstrated on more than one platform.
Why it matters
Below-threshold operation converts quantum computing from a physics argument into an engineering roadmap. If logical error rates fall exponentially with code size, then scaling is a matter of fabricating more qubits with sufficient quality — hard, expensive, and governed by semiconductor-industry learning curves rather than by unknown physics.
It also changes procurement logic. Companies allocating research budgets no longer ask whether fault tolerance is possible; they ask which platform reaches useful logical-qubit counts first, and at what capital cost per logical qubit.
How the platforms differ
Superconducting qubits (Google, IBM) move fast and fabricate like integrated circuits, but each qubit needs its own control wiring, which makes scaling a cryogenics and wiring problem. Trapped ions (Quantinuum, IonQ) have the best fidelities and native connectivity, but gates take microseconds rather than nanoseconds, which limits clock speed.
Neutral atoms (Harvard/QuEra lineage) add reconfigurable connectivity mid-computation, which helps error-correction codes enormously. Topological approaches (Microsoft's) remain the long-game bet: if Majorana-based qubits work as theorized, hardware overhead collapses — but the platform has yet to demonstrate comparable logical performance.
Evidence
The demonstrations are peer-reviewed and reproducible in kind: Google's surface-code experiments showed logical error rates halving with each code-distance increase; Quantinuum and Microsoft reported logical qubits with error rates below physical ones; Harvard's neutral-atom team demonstrated logical operations across dozens of logical qubits in a single machine.
The gap to usefulness is equally documented. Current machines run tens to a few hundred physical qubits at error rates that support small logical circuits. Running Shor's algorithm against RSA-2048 or simulating industrially relevant chemistry requires error rates roughly a trillion times lower than today's logical rates — reachable by scaling, not by discovery.
The competing read
The optimists — largely the platform vendors — project fault-tolerant machines with thousands of logical qubits within the decade, and point to error-correction code breakthroughs (qLDPC codes that need fewer physical qubits per logical one) as overhead reducers.
Skeptics — including several prominent quantum researchers — note that every platform still has a component whose scaling is unproven at the required level: control electronics, laser systems, cryogenic wiring, or fabrication yield. Both camps agree the next milestone to watch is a logical error rate sustained over long computations, not just single rounds.
What happens next
Watch IBM's and Google's published roadmaps for the first machines marketed as error-corrected rather than merely large; watch the qLDPC code race, which could cut hardware overhead by an order of magnitude; and watch for the first commercial workload that runs on a logical — not physical — qubit system. Post-quantum cryptography migration proceeds on its own schedule regardless: defenders should not wait for this timeline.