The Five Ways to Build a Qubit, and Why Nobody Has Won Yet
By Andrew Pyle
The first thing that surprised me about quantum computing is that "quantum computer" is not one thing. It is at least five different machines, built on five different pieces of physics, by different companies, with no agreement on which one wins. That disagreement is not a sign the field is confused. It is a sign the field is early, and it is the single most useful thing for a builder to understand before anything else.
Here is the plain-language map. For each approach: how it works without the math, who is building it, the real strength, and the real weakness. [VERIFY: the company-to-approach assignments below are stable and well established, but confirm current membership before publish, since firms occasionally add a second modality.]
01
1. Superconducting qubits
The qubit is a tiny loop of superconducting metal on a chip, cooled to a hair above absolute zero, where current can flow two ways at once. You control it with microwave pulses. This is the approach most people picture when they hear "quantum chip."
Who: IBM, Google, Rigetti, IQM.
Strength: the gates are fast, and the qubits are printed with chip-fabrication techniques the industry already understands, so scaling the count has moved quickly.
Weakness: the qubits are fragile. They hold their state for a very short time before noise scrambles it, and they need to sit colder than deep space, which is a hard and expensive engineering problem.
02
2. Trapped ions
The qubit is a single charged atom, held in place by electromagnetic fields in a vacuum, and manipulated with lasers. Nature makes every ion of a given element identical, which turns out to matter a lot.
Who: IonQ, Quantinuum, Alpine Quantum Technologies.
Strength: the qubits are extremely high quality. They hold their state far longer than superconducting qubits, they make fewer errors, and any qubit can talk directly to any other, which simplifies real programs.
Weakness: the gates are slower, and trapping and controlling more and more ions with lasers gets hard fast, so scaling the count is the open challenge.
03
3. Neutral atoms
Similar idea to trapped ions, but the atoms are electrically neutral and held in place by tightly focused laser beams called optical tweezers. You can rearrange the atoms in space, almost like pixels.
Who: Atom Computing, Pasqal, QuEra.
Strength: you can hold a very large number of atoms, and the layout is reconfigurable, which opens up approaches the fixed-chip designs cannot easily do.
Weakness: it is a newer approach, and the gate quality is still catching up to the trapped-ion standard.
04
4. Photonic
The qubit is carried by particles of light moving through waveguides and optical components. Instead of trapping something and holding it still, you compute as the light passes through.
Who: Xanadu, PsiQuantum.
Strength: light does not need to be cooled to near absolute zero the way superconducting qubits do, and photons are natural at moving information around, which is attractive for networking quantum machines together.
Weakness: making photons interact on demand is genuinely hard. A lot of the approach is probabilistic, and building reliable deterministic operations is the core difficulty.
05
5. Quantum annealing
This one is the odd one out, and it is important not to confuse it with the rest. An annealer does not run arbitrary programs. It is a special-purpose machine that finds low-energy solutions to optimization problems by letting a physical system settle into its lowest state.
Who: D-Wave.
Strength: it has by far the largest qubit counts available today, and for the specific class of optimization problems it targets, you can use it right now.
Weakness: it is not a universal gate-model computer. You cannot run the famous quantum algorithms on it. It solves a different, narrower kind of problem, and whether that is a durable advantage is still debated.
06
Why nobody has won
The reason there is no winner is that "better" is not one number. A quantum computer is judged on several axes at once, and the approaches trade them against each other:
- How many qubits you have.
- How good each qubit is, measured by how often it makes an error.
- How long a qubit holds its state before noise ruins it.
- Whether any qubit can talk to any other, or only its neighbors.
- How much overhead error correction will demand to make the whole thing reliable.
Superconducting leads on count and gate speed. Trapped ions lead on quality and connectivity. Neutral atoms lead on raw scale. Photonics leads on temperature and networking. Annealing leads on available scale for one narrow job. No single approach leads on all of them, and the one that matters most, error correction overhead, is still being worked out across the board.
07
The builder's takeaway
You do not need to pick a winner, and you should be suspicious of anyone who tells you they already have. The useful move is to learn the axes above, because they are how you read every announcement and every roadmap that follows. When a company reports a big qubit count, ask about fidelity. When it reports record fidelity, ask about count and connectivity. The number they lead with is the axis they are winning, which tells you as much about their weakness as their strength.
The real question is not which physics wins in the abstract. It is which one becomes something you can actually build on first, and that is the question this series is going to keep chasing.
Related
writing
Why Quantum Computers Run Colder Than Deep Space
Noise is the enemy, cold is the defense, and the fridge is half the machine.
writing
Quantum Computing for Builders: A Field Guide
A builder's map of quantum computing, and the honest questions that make the field readable.
writing
Gate-Model vs Annealing: Two Different Machines Both Called Quantum
writing
What a Qubit Actually Is, for People Who Ship Software
writing
Why I'm Going Deep on Quantum Computing (as a Builder, Not a Physicist)
A builder's reason for putting real hours into quantum right now, and what a hands-on, no-hype series will actually cover.