Most quantum hardware news starts with superconducting qubits. IBM, Google, and Rigetti all build them. Yet the qubit itself is easy to misread: it is not a captured atom or a particle, it is a tiny electrical circuit, and understanding it explains both why these machines are so fast and why they need a refrigerator colder than space.
A circuit that pretends to be an atom
A qubit needs two energy levels and a way to drive transitions between them. A simple LC oscillator, an inductor and a capacitor, gives you an infinite ladder of evenly spaced energy levels, which is useless for a qubit because you cannot address two levels without hitting the rest.
A superconducting qubit fixes this by adding a Josephson junction. A Josephson junction is two superconductors separated by a thin insulating barrier, and it acts as a nonlinear inductor. The nonlinearity bends the energy ladder so the first two levels are unevenly spaced. A microwave pulse at that exact frequency drives the 0 to 1 transition while missing the higher levels. The circuit becomes, for practical purposes, an artificial two-level atom.
The transmon
The most common design is the transmon. It uses a large shunt capacitor to reduce sensitivity to charge noise, and it trades away some nonlinearity to get much longer coherence. Nearly every commercial superconducting machine today runs transmons or a close relative. The chip is aluminum or niobium on a sapphire or silicon substrate, printed with the same lithography used for classical chips.
Why 5 GHz and 10 millikelvin
The qubit transition sits near 5 GHz, a frequency chosen because microwave hardware for that band is mature and cheap. The operating temperature, around 10 millikelvin, follows from a different requirement. A qubit must start each computation in its ground state. Thermal noise at 5 GHz would otherwise leave about 1,250 photons in the mode at room temperature. Cooling to 15 millikelvin drops that number below one in ten million. The full cooling path is in What's Inside a Quantum Computer.
Fast gates, short memory
Superconducting qubits run the fastest gates of any commercial modality, tens of nanoseconds per gate, with two-qubit fidelity near 99.9 percent on the best devices. The cost is memory. T1 and T2 times land near 100 microseconds, so a circuit must finish before that window closes. Fast gates and short coherence are the same trade: the qubit interacts strongly with its environment, which speeds control and also speeds decay.
Reading the answer
Readout uses a resonator coupled to the qubit. The qubit state shifts the resonator's frequency, and a microwave probe measures that shift, revealing 0 or 1 without destroying the qubit mid-computation. This dispersive readout is why each qubit carries several control and readout lines, which is why the wiring bundle inside a machine dwarfs the chip.
Why superconducting dominates
Superconducting qubits lead commercial deployment for practical reasons. They print on existing semiconductor fabrication lines, they run at frequencies where off-the-shelf microwave equipment works, and they integrate with the classical control stack. That manufacturing story, more than any single physics advantage, is why IBM, Google, and Rigetti bet the modality. A related result, a superconducting qubit chain cutting logical decay, shows where the hardware is heading.
Next steps
- Types of Quantum Computers Compared: how superconducting stacks against trapped ions, atoms, and photons
- T1 vs T2: what limits the 100 microsecond memory
- What's Inside a Quantum Computer: the fridge that keeps the circuit cold