Term rewriting has been used as a formal model to reason about the complexity of logic, functional , and imperative programs. In contrast to term rewriting, term graph rewriting permits sharing of common sub-expressions, and consequently is able to capture more closely reasonable implementations of rule based languages. However, the automated complexity analysis of term graph rewriting has received little to no attention. With this work, we provide first steps towards overcoming this situation. We present adap-tions of two prominent complexity techniques from term rewriting, viz, the interpretation method and dependency tuples. Our adaptions are non-trivial, in the sense that they can observe not only term but also graph structures, i.e. take sharing into account. In turn, the developed methods allow us to more precisely estimate the runtime complexity of programs where sharing of sub-expressions is essential.