the wiring bottleneck
The wiring bottleneck is the plain, physical reason today's quantum chips can't just keep growing. Most superconducting qubits need a handful of dedicated lines each — typically a drive line to control it, sometimes a flux line to tune it, and a readout path to measure it. So a 100-qubit chip already needs hundreds of cables snaking from room temperature all the way down to the chip at the bottom of the fridge. Multiply that by a million qubits and you simply run out of room, connectors, and cooling.
Each of those lines is a physical thing that costs you. A coaxial cable carries microwave signals, but it also carries heat down from the warmer stages, and the coldest plate of a dilution refrigerator can only remove a tiny trickle of heat — often well under a milliwatt at the base. Every cable also needs a connector, a feedthrough, and space on the chip's edge for a bond pad. Drill holes only get you so far: the perimeter of a chip grows far slower than the number of qubits you'd like to pack in its area, so the wires choke the design long before the qubits do.
This is why people now call wiring, not qubit count, a first-order obstacle to scaling. The proposed escapes are real but early: multiplexed readout shares one line among many qubits; cryo-CMOS tries to move the control electronics down into the cold so signals travel shorter distances; and cleverer cabling and on-chip routing trim the heat budget. None of these is a finished answer yet, and how to wire a truly large machine remains one of the central open engineering questions.
If each qubit needs about one to three dedicated lines, total cabling grows roughly in step with qubit count N — which is exactly why a few thousand qubits already strains a single fridge.
A rough rule of thumb: if each qubit needs its own control and readout lines, scaling is linear in cables — and cable count, heat load, and connector space, not qubit count, are usually what stop a fridge from holding more.