Quantum Computing: Where We Really Are

Quantum computing gets described as "trying every answer at once", which is wrong in a way that makes everything downstream confusing. A quantum computer is a machine that manipulates probability amplitudes so that wrong answers cancel and right answers reinforce. For a small set of problems that trick yields exponential speed-ups. For most problems it yields nothing. Knowing which is which is the whole game.

Qubits, superposition and entanglement

A classical bit is 0 or 1. A qubit is described by two complex numbers — amplitudes — whose squared magnitudes give the probability of measuring 0 or 1. Until measured, the qubit is in a superposition of both. That's not "both at once" in any everyday sense; it's a state with a definite mathematical description that collapses to one outcome when read.

Entanglement links qubits so that their joint state can't be described qubit by qubit. Measure one and the outcomes of the others are correlated in ways no classical system reproduces. With n entangled qubits, the state is described by 2ⁿ amplitudes — which is why simulating even 50 well-entangled qubits exhausts classical supercomputers, and why a quantum computer can, in principle, explore an exponentially large space.

The catch is that you can't read all those amplitudes out. A measurement gives you one classical result. A quantum algorithm has to choreograph interference so that, by the time you measure, the probability mass has concentrated on the answer you want. Only certain problem structures allow that.

What quantum computers are good for

• Simulating quantum systems. Molecules, materials and chemical reactions are quantum-mechanical; simulating them classically scales terribly. This is the most likely first commercially valuable application — battery chemistry, catalysts, drug candidates.
• Factoring and discrete logarithms. Shor's algorithm breaks RSA and elliptic-curve cryptography given a large enough fault-tolerant machine. That machine doesn't exist yet, but the threat is credible enough that the migration to post-quantum cryptography is already underway.
• Unstructured search and some optimisation. Grover's algorithm gives a quadratic (not exponential) speed-up. Useful, but it doesn't turn hard optimisation into easy optimisation, despite what some vendor decks imply.

What quantum computers are not is a faster general-purpose computer. They will not speed up your web server, your database or your neural-network training in any foreseeable form.

Where the hardware really is

Several physical implementations compete: superconducting circuits (IBM, Google), trapped ions (Quantinuum, IonQ), neutral atoms (QuEra, Pasqal), photonics (PsiQuantum, Xanadu) and others. Each trades off gate speed, connectivity, coherence time and manufacturability differently, and no clear winner has emerged.

All share one problem: noise. Physical qubits lose their state in microseconds to milliseconds and every gate introduces error. The solution, quantum error correction, encodes one logical qubit across many physical qubits and continuously detects and fixes errors. The milestone the field has been chasing is a logical qubit whose error rate falls as you add more physical qubits — "below threshold" — and Google's Willow result in late 2024 demonstrated exactly that on a surface code. Since then Quantinuum, IBM and several neutral-atom groups have shown small numbers of logical qubits running real circuits with error rates better than their physical components.

The honest summary as of 2026: we have machines with hundreds to a couple of thousand physical qubits and single-digit to low-double-digit numbers of high-quality logical qubits. Running Shor's algorithm on a cryptographically relevant key would need on the order of a million physical qubits at current error rates. Roadmaps from the major vendors target fault-tolerant systems around the end of the decade; treat those as ambitions rather than schedules.

Getting started as a practitioner

The practical path is hybrid. Classical computers handle orchestration, optimisation loops and everything the quantum processor is bad at; short quantum circuits handle the part with quantum structure. The tooling is mature enough to learn on:

• SDKs: Qiskit (IBM), Cirq (Google), PennyLane (Xanadu) and Amazon Braket's SDK let you build circuits in Python and run them on simulators or real hardware via the cloud.
• Simulators first. Anything up to ~30 qubits runs on a laptop. Build intuition there before paying for hardware time.
• Real hardware for real noise. Running the same circuit on a physical device teaches you more about error mitigation than any textbook.

For most organisations the right posture is informed watching: understand which of your problems have quantum structure (chemistry, some materials and finance problems), track error-correction milestones rather than qubit counts, and start your cryptographic migration now — that part isn't optional.