IBM's Quantum Advantage Claim Challenged by Classical Computation

IBM's claim of achieving 'quantum advantage' has been met with skepticism after a team at Singapore University of Technology and Design (SUTD) computed exact amplitudes for all 2,051 outputs published by IBM in just 37.3 minutes using 256 NVIDIA H100 GPUs.

The experiment, announced on July 30, 2026, used IBM's Boston superconducting processor to run a depth-70 circuit containing 468 T gates on 97 physical qubits. However, the SUTD team used a tensor network contraction method to evaluate all output batches in under an hour.

While IBM's director of research, Jay Gambetta, claimed 'we are now firmly in the quantum advantage era,' the SUTD team's computation is not a direct refutation of IBM's claim. The two teams are measuring different things, with IBM focusing on sampling and the SUTD team evaluating exact amplitudes.

The SUTD team's method uses a contraction width of 35, which is narrower than IBM's estimated 43 using a heuristic path optimizer. This suggests that classical computers may still be able to efficiently simulate certain quantum circuits, undermining IBM's claim of achieving quantum advantage.