The High Speed Machining (HSM) spindle is one of the most critical bearing applications, because it requires both high speed and high power in order to obtain high quality and productivity. Therefore, bearing condition monitoring is important. Firstly, this paper presents a real and typical spindle life example. The vibration signals and their evolution are discussed in relation to the bearing failures that have been observed after the spindle disassembly. Cleavage notably occurred on the ceramic balls of the hybrid ball bearing. Damaged balls and their chippings then damaged uniformly the rings raceways on the whole circumference. As a consequence, a noise component increases in vibration signal due to the worsening of the ball-race contact during the rolling process. In a second section, the noise component produced by bearing condition is studied and characterized. The frequency spectrum distribution is briefly discussed in relation to a signal model. It is demonstrated by Pearson's test that the distribution follows a Gaussian law all along the spindle life. Besides, it evolves with bearing condition. Thus, a new criterion, called SBN (Spindle Bearing Noise), is proposed for the monitoring of the uniformly distributed defect. A specific monitoring device was also developed in order to collect real industrial data during the spindle lifetime. Vibration signals are used in order to evaluate the criterion relevancy by comparison with the current best practices. The analyses through three required conditions for bearing condition monitoring and based on three spindles signals, have shown some good results.