Reversible Computation Models at least do not consume energy, according to physical laws, since no information is erased. And there are also Fredkin Gate and Toffoli Gate which can effective simulate classical irreversible computation circuits. So why don't chips switch to reversible models to save energy consumption? Is it the problem of "noise" that occurs in computation? What kind of "noise" then?
there is a lot of interest however it appears there are some early prototypes but overall not commercial chips so far. see eg Reversible computing is ‘the only way’ to survive Intel's heat mentioning michael frank at UF, his page here RevComp - The Reversible and Quantum Computing Research Group.
from what I can tell it looks like the reversible designs probably require larger chip area to compute the same problems and current designers would rather pay for mainly minimizing chip area at the expense of "heat". and designers have gotten very adept at minimizing heat in nonreversible designs.
one area of new innovation are energy capturing designs that can capture waste heat and convert it back into electricity. eg see Phononic Devices’s Chips Convert Waste Heat into Electricity
the industry is moving toward capturing waste heat for useful purpose eg building heating esp with supercomputers.
of course quantum computation is the "holy grail" for reversible designs but its proving very difficult to create systems due to the issue you mention in qm circuits-- noise or the so called "decoherence problem". the only commercial system so far seems to be dwave which is very expensive & niche right now.