There are currently a plurality of capacity theories of fading channels, including the ergodic capacity for fast fading channels and outage capacity for slow fading channels. However, analyses show that the outage capacity is a misconception. In this paper we use the 1st order Gaussian-Markov process with coherence coefficient $α$ as the unified model for slow and fast fading channels, the capacity of which without channel side information is studied. We demonstrate that the information rate of a fading channel has a structure that the rate of user message is always accompanied by a rate of channel information. The formula for the channel information rate is derived and turns out to be a non-increasing function of $α$. We prove that there is an asymptotically monotonic behavior of the user information rate with respect to $α$ when the input is independent, identically distributed and Gaussian in the high signal to noise ratio regime. It is further conjectured that the monotonic behavior of the user information rate with respect to $α$ is universal.