OpenGL doesn’t know what a GIF is, it just understands textures. If you pack each frame into a single texture, you can utilize sprite batching to reduce draw calls and improve performance. Using a different texture for each frame is not as optimal.
Some desktops support “texture arrays” which could work with GIFs, but at that point you may as well just have an image sequence so you can take advantage of full 32-bit color.
Artists are pretty used to exporting image sequences; tools like Maya and After Effects have it built-in. New versions of AE do not even export to GIF anymore, since nobody is using it these days.
Plus, with texture atlases, you can pack multiple animation sequences into the same image, which should help with filesize and load time. Another thing… with texture atlases you can trim away white space for each frame, leaving you with a much smaller overall image.
Since Java2D uses OpenGL/DirectX under the hood; the more “hardware-friendly” your rendering is, the better it should perform, in theory.