Thoughts exist in the brain, very likely as patterns of transient electrochemical activity.
Although one can imagine various isomorphisms between thoughts and other things (electrical impulses in a computer for instance), until we can actually demonstrate an actual isomorphic process I think we have to withhold judgment as to whether those are thoughts. Details may matter. There certainly isn't anything adequately isomorophic now.
If an idea is an abstraction of a type of thought, then it exists in the sense that a chair exists: particular chairs exist, and one can identify a regularity shared by chairs. But that regularity itself isn't extant; it just tells you how to identify things that are extant. (Thoughts about such a regularity may be extant, just like thoughts about "two" are extant despite there not being a "Two" that exists in the same sense that an apple exists.) One may also wish to identify stored capability to rapidly recall thoughts as "ideas". That seems reasonable, in which case those sorts of ideas act the same way as thoughts: if you have them you have them, and if not you don't.
A potential thought does not exist, just like a potential walk in the park does not. You don't ask "is someone/something in the park because I can imagine walking in the park"? At least, I hope you don't. If you do, pretty soon you're imagining imagining everything and postulating a countable infinity of utterly useless imaginary entities that you wish to say exist. Personally I find this a really poor way to define the concept "exist".
Only when you actually have a thought does it exist. Also, only when you walk in the park are you actually walking in the park.