I use travel data to estimate the consumption value of urban density in the United States. Travel demand offers a transparent way of identifying gains from variety, which are revealed through the decision of individuals to drive farther to a preferred destination. To estimate my model, I use a new micro-geographic database containing all restaurants in an area representing half of the US population, and match each traveler in a large national travel survey with their restaurant choice set. I evaluate at around $70 billion the yearly net welfare gains that individual obtain from the availability of additional eating options beyond the restaurant closest to them. These gains from variety are acquired through 3.5 billion hours of extra travel and accrue disproportionately to individuals living in dense urban areas. City dwellers gain from accessing a destination closer to their ideal, but do not save much on transport costs. This result explains the ineffectiveness of urban density in reducing trip travel time, or, as confirmed by a vast empirical literature, in reducing aggregate vehicle miles traveled.