No form of electricity generation is cheaper today than wind power. However, wind’s natural variability complicates its substitution for conventional coal and gas-powered generation. The planner of an electric grid with a large percentage of wind power would want to know is it always windy somewhere? Here, we investigate the space-time correlation properties of atmospheric winds and their implications for simple electric-grid reliability metrics. Examining coincident wind-speed data from sites separated by distances ranging from a few kilometers to continental scales (5000 km), we find that correlation lengths fall as low-frequency variability is filtered out. Thus, the effective number of uncorrelated wind generators in a given region increases with frequency, and fast fluctuations in generated power are more effectively smoothed by aggregation. Grid outages, desired to be less than “a day per decade,” are determined by the tails of generator outage probability distributions. At least for widely separated wind generators, we find that extreme low-generation probability declines exponentially with increasing number of generators. Thus, our findings favor expanded electric transmission for the integration of wind power into the grid, with benefits of reduced high-frequency wind variability and less frequent low-generation events.