On 3 November 2026 talks and reports said Revolut cut AI coding inference costs by about 90% by switching from commercial tools such as Claude to self-hosted GLM models as the default. An executive said roughly 2,000 engineers are served on 56 Nvidia B200 GPUs at a blended cost near $0.30 per million input/output tokens. The figure is specific to Revolut’s utilisation and tuning, not a portable market price.

What it means: For Greek banks and fintechs with large engineering teams, the case shows AI coding cost can become an infrastructure strategy rather than only a SaaS subscription. Self-hosting needs GPU capital, MLOps, model-licence checks and security rules for code that touches payment systems. Revolut’s savings are not portable; they depend on cluster utilisation and output quality. CIOs should compare commercial-API TCO with private clusters and set rules on which code may come from external models. For Greek businesses and public bodies this means a practical review of contracts, timelines and compliance duties, and it is not investment or legal advice. IT, finance and compliance teams should document owners, review cycles and fallback processes before wider adoption or a vendor change. For Greek businesses and public bodies this means a practical review of contracts, timelines and compliance duties, and it is not investment or legal advice. IT, finance and compliance teams should document owners, review cycles and fallback processes before wider adoption or a vendor change.

Read the original on The Register / AI Engineer talk

Related service: myDATA integrations

Related stories