It can be either. In the demoscene, sceners just call it the scene.
They also like to agglutinate words, e.g. copyparty, sizecoding and so on.
You can split or join the word if you like.
How come we(you?) decided it can be either, given everyone and their dog writes "demoscene" all across the scene?
The demoscene is full of funny-English words — ”compo”, ”graphician”…
I suspect that it, like the "sinus" mentioned elsewhere, may well be a result of the scene (at least at the start) being dominated by people natively speaking Northern European Germanic languages.
I started learning English about 43 years ago, and I still struggle with when words are combined or separate in English because it largely follows no logic while in Norwegian it carries meaning.