European companies might prefer hosting open models on Exoscale's dedicated inference platform.

(I work on ops for that product.)