One of the earliest argument formats philosophy studied was the syllogism, in which two premises yield one conclusion. It could be argued all more complex arguments are made by chaining lots of syllogisms together, making the conclusion of one a premise of another. On this view, everything after the "atomic" premises is a conclusion obtained from them. (By atomic premises, I mean those that are never obtained as conclusions; there may be quite a few of these if we introduce more of them later.)
But how we write an argument is to an extent a matter of orthography. In principle, we can rearrange any argument to have one premise and one conclusion, as long as you admit the conjunction of finitely many statements "counts as" one statement. But we usually think of a long argument as having lots of premises and conclusions, most of them being a combination of the two as we go along. For example, if you read a mathematics textbook that leaves none of its claims' proofs unstated, you could treat the book as proving one conjunction of theorems from one conjunction of axioms, but you wouldn't. You would say, "here's a list of theorems, obtained from this list of axioms (not to mention this list of rules of inference)".
I'll mention one subtlety. Say you're studying a first-order theory with infinitely many axioms comprising a schema that can be summarised as one second-order statement, together with finitely many "standalone" axioms. (This is, for example, what Peano arithmetic & ZF set theory do in order to be first-order.) Then you can't collapse everything you're assuming into one first-order statement, nor everything you derive from it. So sometimes, the way we "count" statements gets into thorny technical aspects.