Quantum machine learning advantage survives real-world noise at 30 qubits

Using simulations of current hardware, researchers tested whether quantum machine learning's theoretical edge survives the noise of real devices — and found it does, at scales as small as 30–40 qubits. The bottleneck for classical rivals isn't computing power but data collection: replicating the noisy quantum protocol with measure-first approaches would take months to years of measurements. By mapping out the impact of gate errors, readout errors, and connectivity constraints, the work suggests this advantage is achievable on near-term quantum hardware today.