We study the proportional dynamics in exchange economies, where each player starts with some amount of money and a good. Every day, players bring one unit of their good and submit bids on goods they like, each good gets allocated in proportion to the bid amounts, and each seller collects the bids received. Then every player updates their bids proportionally to the contribution of each good in their utility.This dynamic models a process of learning how to bid and has been studied in a series of papers on Fisher and production markets, but not in exchange economies. Our main results are as follows:(1) For all linear utilities, the dynamic converges to market equilibrium utilities and allocations, while the bids and prices may cycle. We give a combinatorial characterization of limit cycles for prices and bids.(2) We introduce a lazy version of the dynamic, where players may save money for later, and show this converges in everything: utilities, allocations, and prices.This answers an open question about exchange markets with linear utilities, where tâtonnement does not converge to market equilibria, and no natural process leading to equilibria was known for all additive utilities. We also note this dynamics represents a process where the players exchange goods throughout time (in out-of-equilibrium states), while tâtonnement only explains how exchange happens in the limit. CCS Concepts: • Theory of computation Algorithmic game theory and mechanism design; Convergence and learning in games; Market equilibria; Network games.