Revolut Slashes AI Coding Costs 10x by Moving to Self-Hosted Models
British fintech company Revolut has reduced its AI coding expenses tenfold by migrating away from commercial APIs like Claude in favor of self-hosted open-source models. This architectural shift allows the company to maintain its development velocity while drastically lowering operational expenditures. This move highlights a growing enterprise trend of shifting from expensive proprietary cloud APIs to cost-effective, self-hosted open-weight alternatives for heavy engineering workloads. It demonstrates that organizations can achieve massive financial savings without compromising on the performance required for software development. By deploying open-source models on their own infrastructure, engineering teams avoid the per-token pricing models of commercial providers that scale poorly with high-volume coding tasks. Managing the models internally also grants the fintech firm greater control over data privacy and infrastructure customizability.
## BACKGROUND
Commercial AI APIs offered by companies like Anthropic (Claude) or OpenAI charge users based on the volume of text processed, measured in tokens. Self-hosting involves downloading open-weight models and running them on an organization's own servers or cloud instances using specialized inference frameworks like vLLM or Ollama, offering lower long-term costs for heavy enterprise use cases.