Barclays Report: Cloud Giants Capture $35-$40 of Every $100 Earned by AI Labs
A Barclays research report reveals that major cloud providers (AWS, Azure, and GCP) receive $35 to $40 in inference fees for every $100 earned by AI model companies. This flow of revenue yields operating margins of up to 35% to 45% for the cloud giants. The analysis highlights the unit economics of the generative AI sector, showing that infrastructure providers are major beneficiaries of the AI boom. It also signals a broader industry transition from model training to active inference as commercial adoption grows. AI labs heavily reliant on APIs enjoy higher inference margins (over 80%) compared to subscription-based models (~70%), leading to significant discrepancies in reported gross margins. Additionally, accounting differences in how API revenues are recognized can make direct financial comparisons between AI labs misleading.
## BACKGROUND
AI inference refers to the process where a trained machine learning model runs live data to generate predictions or outputs, which contrasts with the resource-heavy training phase. Agentic workflows are advanced AI-driven processes where autonomous agents make decisions and coordinate tasks with minimal human intervention, driving higher cloud resource consumption.