Barclays: Cloud Giants Capture 35-40% of AI Model Revenue
A Barclays study reveals that for every $100 generated by AI model companies, $35 to $40 flows directly to the three major cloud providers as compute costs.
AI model companies may appear to be generating substantial profits, but the real beneficiaries might be the underlying cloud computing service providers, according to a new study by Barclays. The research breaks down profit distribution within the AI industry, finding that for every $100 in revenue generated by AI model companies, $35 to $40 is paid to Amazon AWS, Microsoft Azure, and Google GCP as inference compute costs.
Cloud providers are reaping significant profits from these payments, earning an operating profit of $10 to $20 on every $100 of AI model company revenue, translating to operating margins of 35% to 45%. This analysis, part of Barclays' "AI Industry Unit Economics" report, offers a clearer view of value distribution across the AI supply chain.
Meanwhile, AI labs themselves are seeing substantial improvements in profit margins. Paid inference profit margins have surged from just over 10% in early 2025 to an estimated 50% to 65% or higher in 2026. Key drivers for this increase include enterprise customer demand and the rise of "agentic workflows," now considered essential market offerings.
Barclays anticipates continued strong growth in the AI sector, but the dynamic between AI companies and cloud providers could shift. By 2028, new financing models and infrastructure projects may begin to challenge the dominance of the current big three cloud providers, potentially leading to changes in market share and profit distribution.