SignalTech/AI
Revolut moves to self-hosted GLM
Revolut adopting self-hosted GLM signals enterprise AI infrastructure shift.
Context
What happened: Revolut moved about 2,000 engineers onto self-hosted GLM for coding and put Claude Code and Codex on an exception list, per David Abitbol, who leads AI ops there.
Key numbers: 56 Nvidia B200s; blended 30 cents per million tokens, roughly a tenth of prior Anthropic and OpenAI spend; engineering lead time
Why it matters: The content says gains appeared once tokens stopped being rationed, with engineers applying AI to all work rather than only hard tickets. It frames open weights as the default and frontier models as the exception, a pricing problem for labs heading to IPO.
What changed
Revolut moved about 2,000 engineers onto self-hosted GLM for coding and put Claude Code and Codex on an exception list, per David Abitbol, who leads AI ops there.
Why it matters
The content says gains appeared once tokens stopped being rationed, with engineers applying AI to all work rather than only hard tickets. It frames open weights as the default and frontier models as the exception, a pricing problem for labs heading to IPO.