The voltage dependence of the low-frequency noise of GaN/AlGaN HEMTs is usually modeled in terms of mobility fluctuations. Noise measurements in commercial devices before and after proton irradiation and/or voltage stress have been found instead to be more consistent with number fluctuations. Low-frequency noise measurements vs. temperature show defect-related peaks in energy at ∼ 0.2 eV, ∼ 0.5 eV, and ∼ 0.7 eV that increase with proton irradiation and/or voltage stress. The Dutta-Horn model of 1/f noise describes these results well, demonstrating that the noise is caused by defects in the AlGaN and/or GaN layers.