Switching (both endogenous and exogenous) are unavoidable recurring features of financial time series, particularly Bitcoin prices. The latter are driven by frequent changes in investors’ sentiment as there is no theoretical fundamental value of crypto assets. Frequent switching in Bitcoin prices—which is often identified in empirical studies—is nevertheless primordially endogenous in nature. Therefore, endogenous switches of such a nature may not be at odds with empirical studies which find that long-memory persistence in economic and financial time series may be spuriously dominated by break points in the data. The theoretical lengths of the switches (i.e. the expected convergence behaviour of shocks to the zero mean) often encounter the intervention of another break point (of an unclaimed length) masking the inherent pattern of the series as being perennially persistent. In the empirical finance literature, both memory (persistence feature) and breaks (types of switching) have implications for the efficiency of the market. In this chapter, we outline an argument and present a theory-driven empirical apparatus to demonstrate that Bitcoin investors are driven, although not beaten by ‘memory’ as well as a number of endogenous switches in the price level. Both memory and endogenous switches are compatible with what we call an endogenous growth mechanism in Bitcoin market. The finding of persistence has also relevance to the theory of learning: an agent that learns synchronously is an agent that will depict fewer persistence behaviours. We employ a variant of the Auto-Regressive Fractionally Integrated Moving Average (ARFIMA) Markov model with endogenous switches to characterise the profile of price fluctuations in cross-market Bitcoin prices. We show that Bitcoin markets depict true long memory and that price dynamics are mostly driven by endogenous feedback mechanisms or market reflexivity.