Hyperscaler AI cloud, neocloud GPU rental, frontier-API hosting — what compute providers are earning from AI. Decomposed into three segments: Frontier lab compute, AI workload compute, and Hosted model APIs.
Q1 2025 → Q1 2026, one line per major seller. 2025 calendar = sum-of-quarterlies basis; Q1 2026 anchored at disclosed run-rates ÷ 4 (MSFT $9.25B = $37B/4; AWS $3.75B = $15B/4 per Jassy). Every major provider grows monotonically into Q1 2026.
ai-index.hepburnadvisory.com.au
iAll Mag3 + Oracle figures are reported on a gross basis (consistent with each company's revenue-recognition policy). Quarterly numbers are pre-Copilot for trajectory continuity; the post-Copilot total is what flows to the headline. Pure-play neoclouds (CoreWeave, Nebius, Lambda, Crusoe) have no Copilot exposure and no token-API pass-through.
Forward commitments
Multi-year contractual commitments reported gross, separately from realised compute revenue. Per the Cross-Ledger Reconciliation rule, these dollars are not rolled into the trajectory chart above — they are forward obligations that flow through over years.
Loading…
iWhy this is separate from the trajectory chart. Google's RPO jump is the contractual order book, not Q1 26 revenue. The Anthropic-Google $200B and Anthropic+OpenAI $718B are forward commitments to MAG3 hyperscalers — they convert into compute revenue over a 5–8 year window, not at signing. Mixing them into a quarterly trajectory would double-count: the contracted dollars already drive the realised quarterly bars over time. Sourced from issuer disclosures of remaining performance obligations and announced multi-year compute commitments.
The Layer-Stack Multiplier
For every $1 of customer Apps Revenue, how many dollars flow down the stack to Compute, Silicon, and Power. The recycling chart nobody else publishes.
Read this chart this way: all four layers on lookback 2025 actuals on the same time-basis (sum-of-quarterlies). Apps Revenue is the measured cohort total from the Revenue Ledger — not editorially extended to ecosystem scale; the cohort-vs-ecosystem gap is flagged on /methodology. Compute is sum-of-quarterlies post-Copilot, net of Hosted model APIs pass-through. Silicon is NVIDIA Data Center FY26 (calendar 2025) + AMD + custom ASIC.
What Would Have to Be True
Convergence signals that would confirm or disconfirm the Compute Revenue trajectory. Each signal has a defined threshold and an observed value.
AI Ledger
Hepburn Advisory
Hepburn Advisory