A Markov Chain is a mathematical structure, not a machine learning architecture. If there are a finite number of states, and a function to determine the probability to transition to any given state from any other state, then it’s a Markov chain.

Transformers with finite block size have a finite number of states, so they are Markov chains.