High Frequency Data and Precision of the mean

I dnt get why high frequency data create more precise variances and covariances, but less precise means. Doesnt high frequency data mean that you collect a lot of data? Isnt that supposed to improve your precision of the mean?