When workers move to a larger city, they tend to experience an increase in earnings. But they also tend to move to larger and better-paying establishments. This paper studies the role of establishment- size composition in explaining the city-size earnings premium. Using administrative data from Spain, we first document a strong positive correlation between city size and establishment size, measured as the number of co-workers. The establishment size for a typical worker is 33 percent larger in a city with twice the population density, even after controlling for worker fixed effects and other observable characteristics. We then decompose the city-size earnings premium into two components: the increase in earnings explained by the increase in establishment size and the within establishment-size premium. We find that 30.8 percent of the short-term gains of moving to a city twice larger can be explained by a transition to a better-paying larger establishment. In contrast, only 5.0 percent of the medium- term gains of accumulating experience in a large city can be attributed to a faster transition to larger establishments. The small contribution to the medium-term gains is due to two facts: first, in large cities establishment size only grows slightly faster than in smaller cities; second, the relationship between earnings and establishment size is weaker in larger cities. Our results indicate that the role of establishment size composition is fundamental for understanding the short-term gains of moving to a larger city but much less so for explaining the medium-term gains.