Apparently the insulation value (R value) of dirt and soil is between .25 and .8 per inch (depending on moisture content). That wouldn’t be great if it was a material like fiberglass, but since it’s dirt cheap (ba-dum-tssh) and easy to pile up in large quantities with little to no ongoing maintenance in this kind of context, it matters.

A 10 ft pile of dirt (assuming 10 ft between heat exchanging pipes and the outside air) has an R value of 24 to 96, which is extremely significant.

I expect there would still be notable losses trying to keep it at 1100F indefinitely, but 10 ft of dirt will have insulation values approximating many feet of fiberglass insulation.

You’d want a very large mass to heat however, scaling matters a lot. You’d want the ratio of surface area to mass to be as small as possible, and that means as large a volume with as thermally dense a material as possible inside. Surface areas increases by the square, while volume increases by the cube.

Also, no matter what you do, you would eventually cook whatever was at the surface or underground, so don’t do this where you want trees - or where there are underground coal seams