Not All AI Revenue Is Created Equal
The AI hype cycle has created a dangerous conflation: investors treating all "AI revenue" as equally valuable. It's not. A company reselling OpenAI API calls with a 20% markup has fundamentally different economics than a company with proprietary models and 85% gross margins. Understanding AI-specific SaaS metrics is the difference between investing in the next NVIDIA and the next WeWork.
The Metrics That Matter
Gross margin: Traditional SaaS benchmarks target 75-85% gross margins. AI companies face a unique challenge: GPU compute costs. Companies running inference on their own models may have gross margins of 50-65%, which looks weak by SaaS standards but is acceptable if margins expand as inference costs decline (they are, at roughly 30% per year).
Net dollar retention (NDR): The gold standard for SaaS health. Best-in-class AI companies show NDR of 130-160%, meaning existing customers spend 30-60% more each year. This happens naturally in AI because usage scales with value — the more a company uses AI, the more use cases it finds.
Token economics: Unique to AI companies. Investors analyze revenue per million tokens processed, cost per token, and how token efficiency improves over time. Companies that can deliver more value per token have structural advantages.
Red Flags to Watch For
API dependency: If a company's entire product is built on OpenAI or Anthropic's API with no proprietary layer, it's a reseller — not a software company. When the foundation model provider adds the same feature natively, the wrapper dies. Check for proprietary fine-tuning, RAG pipelines, or custom models.
Fake AI revenue: Some companies rebrand existing software as "AI-powered" without meaningful AI integration. The tell: ask what changes if you remove the AI component. If the product still works the same, the AI is marketing — not technology.
Unsustainable unit economics: AI companies burning $0.80 to generate $1.00 in revenue are subsidizing growth with investor capital. This works temporarily but creates a reckoning when funding dries up.
Valuation Benchmarks
Revenue multiples by category: Foundation model companies: 50-100x revenue (Anthropic, OpenAI). AI infrastructure: 20-40x revenue (Databricks, Scale AI). Vertical AI SaaS: 15-30x revenue (Harvey, Glean). AI-enhanced legacy SaaS: 10-20x revenue (Salesforce Einstein, ServiceNow). AI wrapper companies: 5-10x revenue — and declining.
Growth-adjusted multiples: The market rewards growth efficiency. A company growing 100% YoY at 60% gross margin gets a higher multiple than one growing 50% at 80% margin. The formula: growth rate + gross margin should exceed 120% (the "Rule of 120" for AI companies, updated from the traditional Rule of 40).
🔒 Protect Your Digital Life: NordVPN
Financial analysis, investment research, and accessing proprietary databases require security. NordVPN encrypts your connections to Bloomberg, PitchBook, and other financial platforms.
Public Market Comps
Use these public companies as benchmarks: Palantir (PLTR) — 25x revenue, 80% gross margins, 20% growth. Snowflake (SNOW) — 18x revenue, 70% margins, 30% growth. CrowdStrike (CRWD) — 20x revenue, 75% margins, 35% growth. NVIDIA (NVDA) — 30x revenue, 75% margins, 90% growth. The private AI companies trading at 50-100x revenue need to grow into public-market multiples within 3-5 years or face painful corrections.
