The AI Valuation Playbook

AI companies report
annualized run rate
and call it ARR

"Any contract less than 12 months in length should be excluded from Annual Recurring Revenue."

ChartMogul, industry standard definition

"So much of this is essentially vibe revenue - money from customers who pay out of curiosity, not because the product solves a persistent problem."

Pat Grady, Sequoia Capital

"Investors wanted to keep evaluating companies as SaaS-predictable, so they tried to shoehorn those elements into 'recurring' revenue. It doesn't truly work."

Nnamdi Okike, 645 Ventures

~20%
of AI "ARR" existed 12 months ago
Only this portion can be tested for recurrence
~55%
of SaaS ARR existed 12 months prior at IPO
More than 2× most AI companies
140%+
NRR disclosed by SaaS at IPO
Applied to revenue that actually existed

For revenue to be "Annual Recurring Revenue," it must have actually recurred. That takes at least 12 months. If a company is 8 months old, 0% of their revenue has been tested. If 80% of their ARR came in the last year, only 20% has faced a renewal cycle.

Traditional SaaS at IPO had ~55% of ARR that existed 12+ months prior, and disclosed 140%+ NRR proving it renewed and expanded. AI companies average ~20% - most of their revenue hasn't had a chance to prove it recurs.

The question isn't what the retention rate might be. It's that most AI "ARR" hasn't even been subject to the test yet.

The Recurrence Test

ARR vs Run-Rate: Reality Check

Only revenue that existed 12 months ago can even be subject to the "recurrence" test.
The rest is run rate. Hover any company to see the breakdown.

Hover over a company to see the breakdown
SaaS at IPO (~55% of ARR existed 12mo prior)
AI Enterprise (50%+ Ent/SMB, LTM rev skew)
AI Consumer-skewed (<50% Ent/SMB)
AI <24 months (minimal renewal data)

Sources

SaaS data from SEC S-1 filings at IPO. AI ARR figures from company announcements, press reports (The Information, Bloomberg, TechCrunch), and investor presentations. ARR 12-months-prior estimated from growth trajectory and funding round disclosures. NRR for AI companies imputed from category benchmarks (ChartMogul, KeyBanc SaaS Survey) - not disclosed.

Test 2: Revenue Composition + Retention Quality

▶

SaaS: uses actual disclosed NRR from S-1 filings. AI: uses segment-based GRR benchmarks. SaaS at IPO beat benchmarks by 40-60pts through expansion revenue - will AI?

Company Deep Dive
Glean Technologies
Enterprise AI search - Founded 2019
$200M
Reported ARR
Retention Scenario
After 12 months
$108M
54% retained
Churn Risk
$92M
46% at risk
Revenue Composition (estimated) ■ Ent ■ SMB ■ Pro ■ Con
Enterprise 85% · SMB 15% · Prosumer 0% · Consumer 0%
Assumed Monthly Churn (based on revenue mix)
5.0% → 46% annual churn
1%
Enterprise
4%
SMB
7%
Prosumer
10%
Consumer
Effective Valuation Multiple
67× 36×
At $7.2B valuation, investors pay 36× on reported ARR. Assuming 46% benchmark churn, they're actually paying 67× on presumed retainable ARR.
What We Know ✓ Verified ◐ Reported ○ Observed