In a cold-plate loop the fluid is mostly water, chosen because nothing else carries heat as well, with glycol added where freezing is possible and an inhibitor package added always. The inhibitors are the actual product: they passivate every metal the fluid touches, and a package matched to copper will not protect an aluminium fitting somebody added later. Biological growth is the other failure path, controlled by biocide and by keeping light and air out of the loop.
Immersion uses fluids that do not conduct. Single-phase systems run on hydrocarbon oils, synthetic esters or similar, which are cheap enough to fill a tank and heavy enough to matter structurally. Two-phase systems need a fluid engineered to boil at a precise temperature just above the components, and the class of fluorinated chemicals that does this is under regulatory restriction as persistent substances — with one major producer having announced its exit from the category. That is a supply question rather than a thermal one, and it has shaped immersion's outlook more than any thermal argument has.