Quantum Outpost

Algorithm Zoo · Optimization

QAOA

Also known as: Quantum Approximate Optimization Algorithm

First described: Farhi, Goldstone, Gutmann, 2014

Status: Heuristic Maturity: Demonstrated Speedup: None proven

The problem

Heuristically find good solutions to combinatorial optimization problems (MaxCut, MaxSAT, ...).

Apply alternating problem-Hamiltonian and mixer-Hamiltonian rotations to |+⟩^n with classical-optimizer-tuned parameters β, γ. p layers; in the p→∞ limit, QAOA recovers adiabatic optimization.

Best classical

Goemans-Williamson (≈0.878 for MaxCut), SDPs, simulated annealing, branch-and-bound. Strong.

Quantum complexity

O(p · m) gates per call for an m-edge graph; classical optimization layer on top.

Our verdict

A research vehicle, not an optimization tool. Ten years of attempts have not produced a QAOA result that beats a strong classical baseline on a problem instance anyone cares about. The interesting research now is understanding *why* — Goemans-Williamson rounding of an SDP appears to be inherently hard to beat with a polynomial-depth quantum circuit.

When to use it

When not to use it

Classical baseline

Goemans-Williamson gives 0.878-approximation for MaxCut in polynomial time. Empirically, modern QAOA at p=10 underperforms simulated annealing on most graph instances; large meta-analyses (Lykov et al. 2023, Bechtold et al. 2025) find no instance class where QAOA wins.

Hardware cost

Per layer: 2|E| two-qubit gates (problem) plus n single-qubit rotations (mixer). At p=10, 20 qubits: ≈400 gates, deep enough that 2026 NISQ noise kills the signal.

Key papers

Deep-dive tutorials

Last verified: 2026-05-24

Weekly dispatch

Quantum, for people who already code.

One serious tutorial per week, plus the industry moves that actually matter. No hype, no hand-waving.

Free. Unsubscribe anytime. We will never sell your email.