We study whether booms and busts in the stock market of the United States (US) drives its volatility. Given this, first, we employ the Multi-Scale Log-Periodic Power Law Singularity Confidence Indicator (MS-LPPLS-CI) approach to identify both positive and negative bubbles in the short-, medium, and long-term. We successfully detect major crashes and rallies during the weekly period from January 1973 to December 2020. Second, we utilize a nonparametric causality-in-quantiles approach to analyze the predictive impact of our bubble indicators on daily data-based weekly realized volatility (RV). This econometric framework allows us to circumvent potential misspecification due to nonlinearity and instability, rendering the results of weak causal influence derived from a linear framework invalid. The MS-LPPLS-CIs reveal strong evidence of predictability for RV over its entire conditional distribution. We observe relatively stronger impacts for the positive bubbles indicators, with our findings being robust to an alternative metric of volatility, namely squared returns, and weekly realized volatilities derived from 5 (RV5)- and 10 (RV10)-minutes interval intraday data. Furthermore, we detect evidence of predictability for RV5 and RV10 of nine other developed and emerging stock markets. In addition, we also find strong evidence of causal feedbacks from RV5 and RV10 on to the MS-LPPLS-CIs of the 10 countries considered. Finally, time-varying connectedness of the RVs of the G7 stock markets is also shown to be strongly (positively) predicted by the connectedness of the six bubbles indicators. Our findings have significant implications for investors and policymakers.