There are 3 main ways to build a QPU, but they’re not competing on the same problem. The question is which architecture fits which job. 1/ Superconducting: $IBM, $GOOGL, $QBTS, $RGTI. Circuits cooled to near absolute zero. Fastest gate speeds in the field. The cost is the dilution fridge: Millikelvin temperatures, extremely expensive helium-3 used for cooling, and exotic wiring. Every added qubit makes the engineering harder. 2/ Trapped ions: $IONQ, $QNT. Single ions held by electric fields. Highest fidelity in the industry (99.99%), and this is what helped it advance to DARPA’s Quantum Benchmarking Initiative Stage C. Also drastically lower manufacturing and cooling costs. But slower gate speeds, because the ions are heavy and moving them takes time. 3/ Neutral atoms: $INFQ. Hundreds of atoms loaded into a laser lattice. Slightly lower two-qubit fidelities (currently 99.9%), but long coherence times. Scaling is a lattice-geometry problem, not a fabrication problem. DARPA shortlisted the 3 architectures because it doesn't know. The winner depends on the workload, not on the physics being better. Superconducting wins if speed matters. Trapped ions win if accuracy matters. Neutral atoms win if you need to scale before 2030. 3 architectures, 3 different risks. Owning one ticker is a bet on a single answer.
View original →
