Barclays: AI Lab Gross Margin Gap Exceeds 50 Percentage Points, Secret Lies in API Mix

In a research note dated August 27, 2026, Barclays pointed out that AI labs' gross margin gaps exceed 50 percentage points due to differences in business structure. The report noted that API inference margins have jumped to 65% in 2026, while subscription product margins are roughly 30 percentage points lower. Under the assumption that both companies generate $100 in total revenue in 2026: the API-heavy player achieves $55 in profit, while the subscription-heavy player with 80% subscription revenue gets only $38 in profit.

The recognition method for indirect API revenue (gross vs. net basis) further amplifies reporting differences. Barclays estimates that for every $100 of AI lab revenue, approximately $35–40 flows to hyperscalers, which generate roughly $10–20 in operating profit (an operating margin of about 35%–45%). The report argues that as inference costs increasingly surpass training costs, AI labs' own margins will improve. However, over the next two years, AWS, Azure, and GCP will still control major compute spending, and only from 2028 onward will gradual infrastructure buildout change the landscape. Barclays reminds investors that when comparing AI labs, they should look beyond revenue figures and dig into underlying variables such as product mix, API vs. subscription share, and revenue recognition methods.

[TechFlow]