AI buildout math: in-service capital earns 121-160% of its cost, well above hurdles
FinanceYF5 · x · 2026-09-23
Exponential View's AI Investment Brief crunches the numbers: the $473bn of AI equipment and data centers already in service carries an annual bill of $145bn (depreciation, 2.4% running costs, 15% return) against $176bn in infrastructure revenue (121% coverage) and $232bn in total AI revenue (160%).
- Future committed capital needs infrastructure revenue to grow just 29% (12-month commitments) to 44% (24-month) annually — well below recent AI revenue CAGR of over 200%; latest compounding is 8.3%, implying a doubling in nine months.
- Two risks: commitments outrunning revenue (watch Oracle's backlog), and revenue quality — supplier-financed purchases and other low-quality instruments now fund 24.6% of the buildout, up from 22.3% in June, on pace to hit alert levels by February 2027.
Related event: Analysis suggests AI infra payback is easier than feared(2 posts)→
More from Venture
- Clio Marks One Year Since $1B vLex Deal, Its Fastest-Growing AI Product Signs a Customer Every 12 Minutes — EJWalters · 2026-11-12
- UBTECH sells 400 U1 Pro humanoids to Yongda Auto in ~$10M deal — CyberRobooo · 2026-11-12
- Physician-founded AI companies on a tear: PathAI up to $1.05B, CounselHealth $625M cash — DrDatta_AIIMS · 2026-11-12
- DeepVest white paper: SEC FY2027 exams target AI tools advisors can't prove they tested — PtrPomorski · 2026-11-12
- YC-backed Sintropix launches AI-native ERP that closes books daily and files taxes worldwide — ycombinator · 2026-11-12
- Polymarket prices 28% odds of a state data center moratorium; Torus raises $500M — Polymarket · 2026-11-12