For twenty-one chapters the qubit has been a tidy arrow in math. Time to ask the blunt question: what is one actually made of? There’s no single answer — a handful of very different physical systems are all racing to be the qubit that scales.
Any system with two clean, controllable energy levels can be a qubit. An atom’s electron sitting in the lower or upper of two orbits. A single particle of light, polarized this way or that. The spin of one electron, up or down. A loop of superconductor carrying current one way or the other. Each turns the abstract |0⟩ and |1⟩ into something you can build, cool, and poke with a laser or a microwave — and each pays for it differently.
Think of racecars built for different tracks. A dragster is blazing fast in a straight line but burns out in seconds; an endurance car is slower per lap but runs for hours. Neither is “best” — it depends what you’re racing. Qubit platforms are the same: superconducting circuits fire gates in nanoseconds but forget in microseconds; trapped ions are sluggish but remember for seconds. What counts isn’t raw speed or raw memory, it’s how many laps fit in the tank.
Pick a platform. The bar chart ranks them by the metric that really matters — roughly how many gate operations fit inside the coherence time (chapter 09’s clock). Notice there’s no clean winner: speed, coherence and scalability pull against each other. Numbers are order-of-magnitude.
However exotic the hardware, it has to satisfy the same five demands — the DiVincenzo criteria — before it can run a single algorithm from Part IV:
Every entry in the chart above is a different bargain struck against this list. None yet aces all five at once — which is precisely why the field still has several horses in the race.
“Whoever has the most qubits wins.” Qubit count is the headline, not the story. A thousand noisy qubits that decohere mid-circuit compute less than fifty pristine ones. The number that matters is error rate per gate — and whether it sits below the threshold where error correction (chapter 10) actually starts helping instead of hurting.
So the real frontier isn’t a qubit headcount — it’s fidelity. Push the per-gate error low enough and a few thousand physical qubits become a handful of flawless logical ones, ready to run the algorithms this whole course was building toward. That hand-off — from fragile physics to reliable computation — is where quantum computing is being won or lost right now.