An indie SaaS founder looked at her 2026 financials and was startled. Her revenue was up 60% year-over-year. Her gross margin had dropped from 87% to 52%. The reason was AI costs that scaled with usage and weren't proportionally offset by pricing increases. She was building more revenue but capturing less of it.

This pattern is now common for indie SaaS that integrates AI heavily. The 30% gross margin floor — below which the business stops being viable for a solo operator — has become a metric worth tracking actively.

Why Margins Have Compressed

AI inference costs scale with usage. Each customer's AI usage costs real money, unlike traditional SaaS where additional users have near-zero marginal cost. Heavy users can be net-negative.

Models are getting more capable and more expensive. As products use frontier models, costs rise. Cheaper models exist but quality trade-offs are real.

Customers expect more AI per dollar. Each year customers expect more capability for the same price. Vendor pricing pressure compresses margins.

Indie operators can't always negotiate enterprise rates. Volume discounts from AI vendors require scale most indies don't have.

Where Margins Leak

Heavy users on flat pricing. A single power user using 50x the AI of typical users on a flat $99/month plan can be cost-negative. Heavy users are common in B2B vertical SaaS.

Free tier with generous AI access. Free tiers that include real AI usage burn money fast. Some indies have killed free tiers entirely.

Premium models on routine work. Defaulting to Opus when Haiku would do. Easy to fix; costly when not addressed.

Failed automation that humans have to redo. When AI work has to be redone by humans, you paid twice — once for the AI, once for the human time.

What Indie Operators Need to Do

Track gross margin actively. Monthly minimum. Per-customer if possible. Knowing where margin is dropping is the prerequisite for fixing it.

Implement usage-based limits or pricing. Either cap AI usage on flat plans or move to usage-based pricing. Both work; doing neither is what kills margin.

Route to cheaper models for routine work. Most agentic workflows can run on Haiku-class models for most steps. The premium models are reserved for the steps that actually need them.

Cache aggressively. Prompt caching, response caching where appropriate, embeddings caching. Each layer of caching reduces vendor spend.

Charge for premium features. Heavy AI features (advanced models, custom fine-tuning, high volume) belong in premium tiers, not in base pricing.

Kill the free tier if it's negative. Some indies have moved to paid-only models with strong success. The free tier was costing more than it was acquiring.

What 30% Floor Means Practically

At 30% gross margin, $10K MRR produces $3K of gross profit. Out of that comes everything else — hosting, tools, the founder's salary. The math gets hard fast.

A healthier indie SaaS in 2026 runs at 55-75% gross margin. Above 75% means you might be under-investing in capabilities (AI usage that would improve customer outcomes). Below 50% means you're at risk.

The discipline isn't optional for indies who want sustainable businesses. The AI cost reality has reshaped indie SaaS unit economics. Founders who watch this metric actively are operating at substantial advantage. Those who don't are running businesses that look healthier on revenue than they are on margin — and the margin reality eventually catches up.