Google Cuts Quantum Computing Costs 11-Fold With AI-Proven Algorithms
Updated
Updated · The Quantum Insider · Nov 10
Google Cuts Quantum Computing Costs 11-Fold With AI-Proven Algorithms
3 articles · Updated · The Quantum Insider · Nov 10
Summary
Google researchers reported AI-designed, formally verified quantum programs that cut combined qubit-and-gate costs by 10.9 times for a fusion simulation, 8.08 times for photochemistry and 2.49 times for a drug-metabolism model.
The system used Lean-based specifications and teams of AI agents that had to prove both correctness and resource use, letting Google optimize programs too large for classical simulation while a separate checker blocked shortcuts that changed the task.
For the fusion case, the improved program dropped to 4,374 logical qubits from 5,650 by changing how particle interactions were represented and by reusing electron-momentum calculations between steps.
The gains still fall short of practical deployment: the paper says many target applications need billions to quadrillions of logical operations, versus roughly 1 million projected for early fault-tolerant quantum processors.
Google said it plans to open-source the Lean library, problem specifications and verified solutions, aiming to let algorithm improvements accumulate alongside advances in quantum hardware.