Skip to main content

Command Palette

Search for a command to run...

AI SaaS Pricing: Variable Costs Kill Margins

Updated
3 min read
AI SaaS Pricing: Variable Costs Kill Margins
D
PhD in Computational Linguistics. I build the operating systems for responsible AI. Founder of First AI Movers, helping companies move from "experimentation" to "governance and scale." Writing about the intersection of code, policy (EU AI Act), and automation.

TL;DR: 73% of SaaS companies underestimate AI costs. Every AI request costs money—unlimited features become margin killers. 4 pricing models that work.

Quick Take: Your "unlimited AI" feature isn't free—it's a ticking cost bomb for your margins. Every AI request has variable costs that traditional SaaS pricing models can't handle. The 73% of SaaS companies underestimating AI operational costs are heading for margin erosion.

73% of SaaS companies offer "unlimited AI" features in their pricing tiers.

They're pricing themselves into bankruptcy.

Unlike traditional software features that scale with infrastructure, every AI request costs money. That innocent "unlimited AI chat" feature becomes a €50,000 monthly bill when power users discover your product.

Why Does Traditional SaaS Pricing Fail for AI Features?

Most SaaS teams treat AI features like any other software capability—build once, scale infinitely. The economics don't work.

Traditional features have fixed development costs and predictable infrastructure scaling. AI features have variable costs per interaction: API calls to OpenAI (€0.03 per 1K tokens), Claude (€0.015 per 1K tokens), or your own inference costs (€0.001-0.01 per request).

Building for clients from dental clinics to FinTech showed me the same pattern: companies that don't account for AI's variable cost structure see 40-60% margin erosion within six months of launching AI features.

The shift required: pricing strategy must align with cost structure, not user expectations.

4 Operational Models That Preserve AI Margins

Bundle with Usage Caps

Include AI features in existing plans but set monthly limits—100 AI queries in Basic, 500 in Pro. Simple for users, predictable for you.

Implementation: Mixpanel for usage tracking (€89/month), Stripe for overage billing. Set caps at 70% of your target cost per user.

The mistake: Setting caps too high. Start conservative—you can always increase limits based on actual usage patterns.

Pure Usage-Based Pricing

Charge per AI interaction: €0.10 per query, €0.50 per document analysis. Aligns costs with revenue perfectly.

Tool options: Lago (€99/month for usage billing), Metronome (€299/month), or custom Stripe integration.

The mistake: Poor communication. Users hate bill shock. Provide real-time usage dashboards and spending alerts.

AI Add-On Packs

Separate AI features into "AI Boost" packages—€29/month for 500 AI credits, €99/month for 2,000 credits.

Psychology: Users who see AI value are willing to pay premium. Clear separation prevents feature dilution.

The mistake: Poor integration. AI features should enhance core workflows, not feel like separate products.

Hybrid Credit System

Base plans include limited AI credits (50-100), with pay-as-you-go for additional usage. "AI Boosters" for temporary capacity increases.

Implementation: Credit system in your database, automated top-ups via Stripe. Track usage with unified dashboards showing cost per user, feature adoption, and margin impact.

The Implementation Sequence

Week 1: Audit current AI usage patterns and costs per user. Install usage tracking (Mixpanel or custom analytics).

Week 2: Model pricing scenarios—calculate break-even points for each pricing approach.

Week 3: A/B test with 10% of new users. Monitor conversion rates and usage patterns.

Expected outcome: 15-25% improvement in unit economics within 60 days, with clearer path to profitability as AI usage scales.

The Diagnostic Question

Pull your last month's AI infrastructure costs and divide by active AI users. If that number is above 30% of your average revenue per user, your pricing strategy needs immediate attention.

Most SaaS teams discover they're subsidizing power users while casual users pay the same price. The gap between cost and revenue reveals which pricing model fits your usage distribution.

If this resonated, these will sharpen your perspective:


Originally published by First AI Movers on LinkedIn. Written by Dr Hernani Costa, Founder and CEO of First AI Movers.

Subscribe to First AI Movers for daily AI insights and practical automation strategies for EU SME leaders. First AI Movers is part of Core Ventures.

Ready to automate your business? Book a call today!

1 views
AI SaaS Pricing: Variable Costs Kill Margins