ARR
Annualized Recurring Revenue · take current monthly recurring revenue × 12. The standard AI/SaaS revenue metric.
Annualized Recurring Revenue · take current monthly recurring revenue × 12. The standard AI/SaaS revenue metric.
Basic
ARR is not GAAP revenue. It's the run-rate of current subscription revenue, annualized. If a company has $10M MRR, they report $120M ARR. Investors and analysts track ARR because it normalizes subscription revenue timing and shows growth velocity. For AI companies: OpenAI $10-15B ARR (est), Anthropic $5-8B ARR (est), Cursor $500M+ ARR.
Deep
ARR methodology varies: some companies include one-time revenue (contract bookings ÷ contract length), some include usage-based. Pure SaaS ARR = MRR × 12 from recurring contracts only. Most AI companies include usage-based revenue because pure subscription misses the heavy API users. This makes AI ARR growth rates look explosive (OpenAI reportedly grew from $1.6B ARR in late 2023 to $10B+ in late 2024) but the underlying economics are usage-dependent, which is less predictable than SaaS subscription. Churn in usage-based is dangerous · a single enterprise customer dropping can wipe $10M ARR overnight.
Expert
AI ARR is structurally different from SaaS ARR: usage-based revenue is more volatile, growth rates are inflated by token cost deflation (more usage per dollar as prices drop), and revenue concentration risk is higher (top 10 customers often 50%+ of ARR). The Rule of 40 (growth rate + margin should exceed 40%) is hard to apply because AI COGS are high and improving fast. AI gross margins in 2026: 30-55% for API providers, 60-80% for UI/wrapper apps. Frontier labs operate at loss when all costs are included.
Depending on why you're here
- ·ARR = MRR × 12 for pure SaaS, includes usage for AI
- ·Volatility >> SaaS · usage-based revenue responds to customer workload changes
- ·Revenue concentration typically 50%+ in top 10 customers
- ·ARR comparisons between companies are fraught · definition varies
- ·Watch underlying revenue growth separately from "ARR" headlines
- ·AI ARR rides on workload adoption, not customer count
- ·AI ARR growth rates (5-10× YoY) are unprecedented · but fragile
- ·Rule-of-40 hard to apply · COGS are high and model-dependent
- ·Churn in usage-based revenue is less visible but just as dangerous as SaaS logo churn
- ·How much a company is making per year, right now · not future predictions
- ·Different from revenue you see in financial statements
- ·The headline number you see in "X reaches $Y billion ARR" stories
AI ARR is inflated by token-cost deflation · more tokens per dollar means more revenue with the same underlying customer workload. Watch active workload growth, not ARR.