Most qubits are fragile because they are simple: a single spin or a single photon state, knocked over by the faintest stray vibration or stray photon. The French firm Alice & Bob is betting on a stranger unit, the "cat qubit," named for Schrodinger's cat because it holds two opposite states at once, but engineered so that one of the two dominant error types, bit-flips, is suppressed by the physics of the device itself rather than by error-correction code layered on top. France's government-backed quantum plan is now funding Alice & Bob to wire that hardware directly into a national supercomputer, so a classical machine can handle the outer loop of a calculation while the cat qubits handle the part that needs quantum weirdness. It is the same logic behind the U.S. Department of Energy's Quantum Genesis Q Competition, a $215 million program announced September 17 to fund private teams building fault-tolerant machines: nobody thinks more raw qubits alone gets there, so the money is chasing designs where errors are cheaper to fix in the first place.

Fujitsu's approach, a prototype disclosed this month, picks a different physical trick: it uses tin-vacancy defects in diamond, single missing atoms replaced by a tin atom, as the qubit itself, then links them into photonic circuits so they can talk to each other over light rather than fragile direct wiring. The pattern across all three efforts, France's cat qubits, Fujitsu's diamond defects, and the DOE's open call, is that the fault-tolerance race has stopped being about qubit count and started being about which physical substrate makes errors rarer before correction even begins. Fujitsu has not published a qubit count or a gate fidelity number, and until it does, there is no way to weigh its diamond approach against Google's, IBM's or Quantinuum's processors, all of which the QPU-BM benchmark showed trailing classical simulation by a factor of 100,000 as of September 15.