AI-specific revenue by buyer segment, channel, provider, and cost outcome.
Revenue Flow
Scroll horizontally to explore the Sankey, or rotate to landscape.
Provenance tier:
1 Sourced (A = first-party disclosure, B = corroborating sources)
2 Derived (A = deterministic calc from Tier 1, B = triangulated)
3 Projected (A = anchored extrapolation, B = interpolated, C = scenario assumption)
4 Editorial (replace ASAP)
— hover any node for source citation
?
Why do the middle columns shrink? Each column shows the dollars that flow through it. Channels (Hyperscalers, Trad. SaaS) retain a margin before passing revenue to Model Providers — Hyperscalers clip ~20%, Trad. SaaS ~60%. That retained margin flows directly to Generated Cashflow (bottom right), bypassing Model Providers entirely. So each column is smaller than the one before it — the drop is the channel margin.
i
Per-archetype channel routing. Each provider's revenue routes to channels per its entity archetype — frontier labs send ~95% of API revenue direct (5% via hyperscaler resale); AI Natives ~90% direct; Enterprise SaaS ~50/50. The Who Pays column splits AI buyers into AI Natives (heavy API consumers like Cursor / Glean / Perplexity) and Enterprises & Govs (Fortune 500, regulated, sovereign). See methodology § entity archetype taxonomy.
!
Scope. The Revenue Sankey shows model providers earning AI-attributable customer revenue in the year. Labs that don't directly monetise their models — Meta (Llama, open-weight, ad-funded indirectly) is the canonical example — are excluded. Their AI spend appears on the Capital Ledger (capex) and Compute Ledger (chip purchases) instead.
Click or hover a node to see flow details and tier provenance