A qubit is an abstraction — "a two-level quantum system." The engineering question is which physical system, and there is no single answer. Several very different technologies each realise a qubit, and which will win (or whether several coexist) is genuinely open. This is what the race actually looks like.
The main approaches
- Superconducting qubits — tiny circuits cooled to near absolute zero, where electrical current flows without resistance and behaves quantum-mechanically. Gates are extremely fast, and the fabrication borrows from the chip industry. This is the path favoured by several of the largest efforts. The downside is short coherence times and the need for dilution refrigerators.
- Trapped ions — individual charged atoms held in place by electromagnetic fields and manipulated with lasers. They have excellent coherence and fidelity, and every qubit can interact with every other — but gates are slower and scaling the trap is hard.
- Photonic qubits — information carried by particles of light; naturally robust against decoherence and well-suited to networking, but two-photon gates are difficult.
- Neutral atoms, spin qubits in silicon, and topological qubits (which aim to be error-resistant by physics) are all being pursued.
Each embodies the same qubit maths from Foundations; they differ in the messy practical dimensions — how long the state survives, how accurate each gate is, how many qubits you can connect, and how you scale up.
The NISQ era
We live in what physicist John Preskill named the NISQ era — Noisy Intermediate-Scale Quantum. Machines have tens to a few thousand physical qubits: "intermediate-scale". But they are "noisy" — errors accumulate faster than today's partial error correction can fix, so you can only run short circuits before the result dissolves into noise. NISQ machines are real, programmable and scientifically valuable, but they cannot yet run Shor's algorithm on anything of cryptographic size.
Reading the hype
A practical expert habit: qubit count alone is close to meaningless. A machine with 1,000 noisy, barely-connected qubits may be less capable than one with 100 high-fidelity, fully-connected ones. What actually matters is coherence time (how long qubits survive), gate fidelity (how accurate each operation is), connectivity (which qubits can interact), and ultimately how many logical (error-corrected) qubits emerge. When you see a headline touting a qubit number, ask about those four — it's the difference between a real advance and a press release.
With how they're built understood, the final lesson steps back: what will quantum computing actually be good for — and what it won't.