One of the problems is that a normal file format doesn’t work for a very very large image.
The orders of magnitude difference simply don’t work in your favor.
What tools like Google Earth (and other GIS tools) do, is store images in multiple resolutions.
Start with an overview image at some reasonable resolution, like 8kx8k.
Then, split the image into four quadrants, and render each of those quadrants in 8kx8k.
Then, keep splitting each of those quadrants, and so forth, until you get the resolution you need at the bottom of the “pyramid” of images.
When displaying a large overview image, you start with the top-level image in the pyramid.
As you zoom in, you go “down” in the pyramid and include sub-images that overlap the view area, until you get to the point where image resolution is adequate for the screen (no need to go further down the pyramid.)
You might thing this uses a lot more space than just a single, very large image, but it’s actually only about 30% bigger, because of the law of exponential functions.
(This is very similar to MIP maps, btw, except you do the tiling of each individual layer as an add-on.)