In several interesting markets, demand is an increasing function of past sales because of learning, network externalities, or fashion. This paper examines entry into such markets. The two key elements of the model are that firms are uncertain about the demand (and learn in a Bayesian fashion) and that demand grows endogenously over time. The capacity expansion path of the competitive market is compared with the planning/monopoly solution. These paths differ not only with respect to levels (the market's investment is too low), but also their time patterns (externalities may lead to S-shaped diffusion). This framework provides some justification for industrial or trade policy arguments for subsidizing entry into new markets, especially for infant-export industries. The markets examined also exhibit path-dependence: small initial differences in demand conditions may lead either to an established market or a non-existing one.