This work introduces a time-scale-based methodology for predicting the lifetime of ultra-thin gate oxides by analyzing gate leakage current under low-voltage stress conditions. By leveraging a baseline construction algorithm and a physics-informed machine learning framework, the approach identifies individual defect percolation paths from gate leakage current data collected at 1.2 V across 130 high-𝜅 n-type metal–oxide–semiconductor capacitors. Analysis of the amplitudes and first-occurrence times of individual leakage paths reveals distinct regimes associated with process-induced and stress-generated defects. The extracted defect generation exponent from a power-law model facilitates prediction of time-to-soft breakdown value, which align closely with the extrapolated value from traditional high-voltage stress methods. These findings validate high-voltage measurements as a fast reliability extraction method since the generation mechanisms at low voltage can indeed be extrapolated from high-voltage stress conditions.