I examine entry into markets where demand is an increasing function of past sales because of learning, networks, or fashion. Demand is initially unknown (with firms learning in Bayesian fashion) and grows endogenously over time. The competitive expansion path and the efficient/monopoly solution differ not only with respect to levels (the market’s investment is too low), but also time patterns: externalities contribute to S-shaped diffusion. There is also path-dependence: small initial differences may determine whether the market will grow or not open. Policy arguments for subsidizing entry into new markets, especially in infant export industries, are examined.