devrandom comments on We are headed into an extreme compute overhang

[devrandom]

New Transformer specific chips from Etched are in the works. This might make inference even cheaper compared to compute.