A new methodology is proposed to estimate the rotating phase of a brushless direct current (BLDC) motor under variable-speed condition through phase current analysis. Studies on this topic are limited. By adopting several signal processing techniques, including zero-phase filtering, Hilbert transform-based phase estimation, and signal truncation and alignment, an accurate rotating angle curve can be obtained from the noisy current signal featured with frequency and amplitude modulations. Particularly, a criterion called sinusoid similarity is proposed to evaluate the error of phase estimation and to guide optimal filter design. The proposed methodology is then combined with the vibration signal analysis-based order tracking technique to achieve variable-speed motor bearing fault diagnosis. Experimental results indicate that the proposed methodology provides a simple, noninvasive, highly accurate solution for the estimation of BLDC motor rotating phase, thereby presenting potential applications in motor bearing fault diagnosis and other related areas.