multi-chip module (MCM)

A multi-chip module is a way to build one big quantum processor out of several small ones. Instead of fabricating every qubit on a single large die and hoping it all comes out perfect, you make a handful of smaller qubit chiplets, test each one on its own, throw away the duds, and mount only the good survivors side by side on a shared carrier so they wire up and act as a single machine. Picture building a wall from pre-inspected bricks rather than pouring one enormous slab and praying it has no cracks.

The reason this helps comes down to yield. On a single big die, one bad qubit, one lossy junction, or one element landing at the wrong frequency can spoil the whole chip, and the chance of at least one defect climbs fast as the die grows. Split the design into chiplets and a failure costs you only that small tile, not the entire processor — and you can pre-screen tiles for the right frequencies and coherence before committing them. The tiles sit on a common carrier or interposer that routes signals between them, usually joined by flip-chip bumps or short superconducting links, so a control line entering the module can reach a qubit several tiles away.

The honest catch is that the seams are hard. A chip-to-chip junction adds a little loss, a little reflection, and a mechanical joint that has to survive being cooled to a fraction of a degree above absolute zero, and a coupling that crosses between tiles is almost always weaker and noisier than one inside a tile. So an MCM trades a yield problem for an interconnect problem: real modules exist and are an active path to scaling, but stitching tiles together without quietly degrading the qubits at the boundary is still very much unsolved engineering.