The paper shows that US GDP velocity of M1 money has exhibited long cycles around a 1.25% per year upward trend, during the 1919-2004 period. It explains the velocity cycles through shocks constructed from a DSGE model and annual time series data (Ingram et al., 1994). Model velocity is stable along the balanced growth path, which features endogenous growth and decentralized banking that produces exchange credit. Positive shocks to credit productivity and money supply increase velocity, as money demand falls, while a positive goods productivity shock raises temporary output and velocity. The paper explains such velocity volatility at both business cycle and long run frequencies. With filtered velocity turning negative, starting during the 1930s and the 1987 crashes, and again around 2003, results suggest that the money and credit shocks appear to be more important for velocity during less stable times and the goods productivity shock more important during stable times.