• CQuIC Seminars

July 16, 2025 3:30 PM - July 16, 2025 4:30 PM
PAIS 2540

Host:

Ivan Deutsch

Presenter:

Dr. Michael Perlin (JPMorgan Chase)

Zoom link

Optimization is often cited as a promising application of quantum computers.  However, the low degree of provable quantum speedups has led prior rigorous end-to-end resource analyses to conclude that a quantum computer is unlikely to surpass classical state-of-the-art on optimization problems under realistic assumptions.  In this work, we compile and analyze the Quantum Approximate Optimization Algorithm (QAOA) combined with Amplitude Amplification (AA) applied to random 8-SAT at the satisfiability threshold.  Our compilation involves careful optimization of circuits for Hamiltonian simulation, which may be of independent interest.  We use the analytical scaling of the time-to-solution for QAOA identified by Boulebnane and Montanaro in PRXQuantum.5.030348, and find that with QAOA depth p=623, QAOA+AA achieves a crossover with state-of-the-art classical heuristics at 179 variables and 15 hours of runtime when executed on a surface-code-based fault-tolerant quantum computer with 74 million physical qubits, a physical error rate of 1E-3, and a 1 microsecond code cycle time.  Notably, we allow the classical solver to be parallelized as long as its total energy consumption is equal to that required for decoding in the surface code.  We further show that this restriction on classical solver energy consumption can be relaxed given optimistic but plausible reductions in physical error rates and fault-tolerance overheads, enabling a crossover of 3 hours using 9 million physical qubits against a classical solver running on a supercomputer with 725, 760 CPU cores.  These findings support the hypothesis that large-scale fault-tolerant quantum computers will be useful for optimization.

*Zoom Password Available Upon Request: Contact ashlosberg AT unm.edu