Revolut dropped Claude and cut AI inference costs by 90%
Revolut has cut its AI inference costs by 90% in less than a year by ditching Anthropic’s Claude and moving to self-hosted, open-weight models from China’s Z.ai.
The London-based digital bank had previously tested Google’s Gemini before settling on Claude. The problem: heavy users were spending thousands of dollars per month on tokens without generating equivalent ROI. Revolut Infrastructure Lead David Abitbol described the internal mood plainly at a recent conference:
“We were highly skeptical about the cost of AI. When we launched it we launched it with strict budget limits.”
What the switch looks like in practice
Revolut moved its workloads to Z.ai’s GLM models, running them on its own infrastructure. GLM-5.1 brought inference costs down to around $0.30 per million input/output tokens, roughly one-tenth the cost of comparable commercial options. The bank now runs 56 Nvidia B200 GPUs. The model alone requires around 380GB of GPU memory, plus an estimated 6 to 26GB per user for context retention.
Speaking at GTC Berlin, Abitbol reported a 70% improvement in lead time productivity and a 60% improvement in coding time productivity after the transition. He also cited data privacy and auditability as reasons for choosing Z.ai’s GLM models specifically.
The tradeoff worth naming
Self-hosting is not free. GPU hardware, electricity, and ongoing maintenance are real costs that offset the per-token savings. Revolut is large enough to absorb that capital outlay. The benefit they gained beyond raw cost reduction was predictability: no token limits, no surprise bills at month end.
Abitbol put it directly:
“Open source is now a priority at Revolut, and our economics are favorable. The billing is flat, we don’t have any bill shock at the end of the month, and the frontier is the exception.”
The operator takeaway
For most solopreneurs and small teams, 56 B200 GPUs are not a realistic option. But the underlying logic transfers: using a cheaper, purpose-fit model for routine coding and operational tasks, rather than defaulting to a frontier model for everything, can dramatically change your AI cost structure. The unpredictability of commercial API billing is a real pressure point, and Revolut’s move is a case study in what it looks like to solve that problem at scale.
I’m founder of BizStack at Cagri Sarigoz LLC and a passionate advocate for entrepreneurs.
With over 14 years in tech, marketing, and AI, including my role as Head of SEO at CitizenShipper and co-founder of TaleBot at Intale AI, I’m dedicated to sharing genuine, useful product insights and tips.
At BizStack, I aim to cut through the digital noise to provide clear, actionable advice.
And more than all else, I’m a father to a (always) little girl and a husband.
Contact me at [email protected] for assistance.
