Revolut cuts AI coding costs 10x by self-hosting GLM

In early November 2026, Revolut revealed it cut its AI coding inference costs tenfold in under a year by making self-hosted GLM models the default for roughly 2,000 engineers, served on 56 NVIDIA B200 GPUs. It is the first hard-number proof that self-hosting an open-weight model pays off — but only at a certain scale.