How to price your AI SaaS in 2026 (without giving away the margin)
Most solo founders price their AI SaaS by guessing. The result: either paying customers leave when they realize the value, or the founder discovers their unit economics are negative because every user costs more in OpenAI fees than they pay. Here is the actual framework that works.
You're about to ship pricing. You've been staring at competitor prices, ChatGPT-asking for advice, and bouncing between $19/mo and $99/mo for two weeks. Someone on Twitter said "just charge more." Someone else said "free forever." Neither helped.
This is the actual pricing framework for AI SaaS in 2026. Built around the constraint nobody talks about: AI SaaS has variable cost per user that classical SaaS does not, and most pricing advice assumes classical SaaS.
TL;DR
1. Start with your variable cost per user (LLM API calls + storage + bandwidth + payment processor + auth + monitoring + security tools). 2. Multiply by 5-10x for your gross margin target. 3. Sanity-check against what your specific persona will actually pay. 4. Pick a pricing model (per-seat, per-usage, tiered, hybrid) — usually tiered. 5. Lock the pricing for 6+ months and iterate based on real data.
If you skip step 1, you might price below your variable cost and lose money on every user. The AI SaaS pricing trap.
The cost-floor calculation nobody does
Your AI SaaS has variable cost per user. For a typical chatbot SaaS:
- LLM API calls: $0.20-2.00/user/month (depending on chat volume + model)
- Vector embedding storage: $0.05-0.30/user/month (Pinecone / Qdrant / pgvector)
- Bandwidth: $0.10-0.50/user/month (depends on UI weight)
- Payment processor (Stripe): 3% + $0.30 per charge
- Auth (Clerk): $0.02/MAU past free tier
- Monitoring (Sentry): negligible per user
- Security (Securie): $0/month (Free) to $0.10/MAU at Solo Founder tier
Total: typically $0.50-3.50/user/month in variable cost. Some apps (heavy LLM use) exceed $10/user/month.
If you charge $9/month and your variable cost is $4/user/month, your gross margin is 56%. SaaS gross margins of 75-90% are the standard for healthy businesses; below that, you cannot afford support, marketing, or development.
The mistake solo founders make: pricing based on competitor prices without checking their own variable cost. If your competitor is venture-funded and burning money, their $9/mo pricing is a customer-acquisition strategy that doesn't reflect actual unit economics. Following them puts you out of business.
The 5-10x multiplier rule
Once you know your variable cost:
- 5x is the absolute floor for gross margin (covers support, dev, marketing, but tight)
- 8-10x is healthy (room to grow + invest)
- 15x+ is the venture-scale ratio (enables aggressive marketing + free tiers)
If your variable cost is $1.50/user/month, you should charge minimum $7.50/month, comfortably $12-15/month, aggressively $20+/month if your value prop justifies it.
This is the floor. The ceiling is what your persona will actually pay.
What your persona will actually pay
Vibe-coder solo founders have specific psychological price ceilings in 2026:
- $0: their default expectation for trying anything
- $15-20/month: cursor-class psychological ceiling. Most consumer-grade tools (Cursor, Linear, Notion Personal, Lovable hobby) cluster here.
- $49/month: "this is a real tool" tier. Implies the user is spending on their actual business.
- $99/month: "we're a real company" tier. Founder pays from a business credit card.
- $299/month: small-business or seed-funded tier. SOC 2 / SSO / audit-log surfaces become reasonable to pitch.
- $500+/month: enterprise pre-conversation. Requires sales conversation, contract, often a procurement process.
The psychological ceiling is a real ceiling. A $19/mo tool that should be $29/mo will get more revenue than a $35/mo tool with the same value. The crossing of $20 is where the user goes from "instant decision" to "let me think about it."
The flip side: above $99/month, the user expects sales involvement. A self-serve $149/month tier rarely converts because the user expects a salesperson to validate the purchase.
The four-tier pattern that works
For AI SaaS targeting solo founders + small teams:
| Tier | Price | Who buys | What they get | |------|-------|----------|---------------| | Free | $0 | Anyone trying it | Limited usage, watermark, fewer features | | Starter | $15-19 | Solo founders | Reasonable usage, core features, no enterprise | | Pro / Team | $49-79 | Small teams (2-10 people) | Higher usage, collaboration, integrations | | Business | $199-299 | Growth-stage / first-paying-enterprise | Higher usage, SOC 2 evidence, SSO, audit logs | | Enterprise | Custom (typically $500+) | Sales-led | Custom contract, SAML, dedicated support |
This pattern fits ~80% of AI SaaS targeting the vibe-coder + solo-founder persona. The numbers shift by category (developer tools tend higher; consumer-creative tools lower) but the structure holds.
The key choices: - Free tier vs no free tier: free tier is acquisition; pure-paid is monetization. Pick based on which bottleneck you face first. Most solo founders need acquisition; free tier wins. - Per-seat vs per-usage: per-seat is predictable revenue; per-usage matches cost-of-service. AI SaaS often uses hybrid (seat with usage cap, overage charges OR throttle). - Annual discount: 15-20% off for annual is standard. Predictable revenue + reduces churn.
Per-usage vs per-seat — the AI SaaS dilemma
Classical SaaS prices per seat — every user costs ~the same to serve. AI SaaS doesn't work that way: a heavy user can cost 100x what a light user costs, and seat-based pricing means you're losing money on heavy users while overcharging light ones.
Three approaches:
### 1. Pure per-seat with usage caps
Charge per user, cap usage at a reasonable level for that tier (e.g., 100 messages/month on Starter, 1,000 on Pro). Users above the cap upgrade or get throttled.
Pros: predictable revenue, easy to explain, matches user mental model. Cons: heavy users hit the cap and churn; light users feel they're overpaying.
### 2. Hybrid (seat + usage with overage)
Charge per user with a generous included usage allowance, then charge per-unit for overage.
Pros: captures heavy-user revenue without churning them, light users feel fair. Cons: harder to explain, harder to predict revenue, billing complexity.
### 3. Pure per-usage
Charge per request / per token / per generation. No seat charge.
Pros: matches cost-of-service exactly, customers feel they pay for what they use. Cons: unpredictable bills produce churn, harder to model unit economics.
For most AI SaaS targeting solo founders: hybrid (seat with cap, throttle past cap) works best. The Cursor / GitHub Copilot pattern.
Capped-envelope pricing — the model that survives bill-shock
The bill-shock problem: a user with a leaked API key, a runaway script, or just an unexpectedly heavy month sees a $4,000 bill instead of their normal $20. They churn. They tell everyone on Twitter.
Capped-envelope pricing solves this. Users pick a tier with a soft cap; usage past the soft cap throttles instead of charges. The bill is predictable; the user is never surprised.
The implementation: cost-firewall middleware that throttles when per-user usage crosses the soft cap. Securie's L39 cost-firewall is built for exactly this — your app inherits the cost-firewall behavior for free.
The trade: you cap your upside (heavy users don't pay more, just get throttled). For most consumer SaaS that's fine; the predictability is the feature.
Pricing testing — the right way
Once your initial pricing is live, you'll want to test changes. The honest answer about pricing testing:
- Don't A/B test pricing across users. It feels like a good idea; it produces customer-trust collapse when users see different prices on Twitter.
- DO change pricing for new signups while grandfathering existing customers. Standard practice; users accept it.
- Watch for: signup conversion rate at each tier, upgrade conversion rate from free → paid, churn rate at each tier, MRR growth.
- The signal that pricing is wrong: nobody upgrades from free to paid (free tier too generous OR paid tier too expensive); OR everyone upgrades immediately (free tier too limited; you're leaving money on the table).
Wait at least 6 months between major pricing changes. Pricing-change-fatigue is a real customer-trust cost.
What changes when you raise prices
The right pricing change at the right time signals confidence + maturity. The wrong pricing change at the wrong time signals desperation.
Right time to raise: you have 50+ paying customers, your churn is under 5% monthly, your customers are getting genuine value (your support volume and feature requests reflect a healthy product). Raise by 30-50%; grandfather existing customers; explain the change in a brief email.
Wrong time: pre-PMF (you don't know what value you're delivering yet), or right after a quality issue (your customers will read the price increase as you trying to extract more from a worse experience).
What changes when you lower prices
Almost never the right move. Lowering prices signals desperation, churns the customers who valued you at the higher price, and rarely produces enough new customer volume to make up for the unit-economics hit.
Exceptions: launching a lower-tier product (intro pricing for a new SKU), or correcting a clearly-wrong initial pricing. In both cases, frame as "we added a new tier" rather than "we lowered prices."
Where Securie fits in your pricing math
If your pricing target is $49/month per user (Solo Founder tier), your variable-cost target is below $5/user/month. The cost components:
- Securie security on every PR + per-user attestation: ~$0-0.10/user/month
- Cost-firewall ledger preventing $4,200 OpenAI bill: ~$0/user/month (it's protecting you, not costing extra)
- Production-Readiness Cert your prospects can verify: ~$0/user/month (included in tier)
The Securie cost is in the noise. The Securie value is preventing the $4,200 bill, the launch-day data leak, and the prospect-trust collapse from "is your app secure?"
Try Securie — free during early access. The Solo Founder tier ($49/mo) maps cleanly onto your own $49/mo pricing.
Related
Related posts
If you're a solo founder budgeting for your first AI-built SaaS, this is the honest cost breakdown — every line item, every free tier, every gotcha that turns a $50/month plan into a $1,200 surprise. Written for the moment before you pick your stack.
It's 3 AM. You scrolled X and saw a tweet about a Lovable / Bolt / v0 app leaking customer data. You start wondering if yours is next. Here is the exact checklist to run in the next 30 minutes — what to check, what to fix first, and how to stop having this problem.
A prospect just emailed asking 'is your app secure?' You don't have a real answer. Here is the honest playbook — what to say, what evidence to point at, and how to turn this question from a deal-stopper into a deal-accelerator. Written for solo founders who don't want to lie.
Every AI-generated Next.js app ships with middleware.ts that looks like it gates admin routes. Half of them do not actually run on the routes they think they run on. Here is the 5-minute test, the canonical bugs, and the fixes — written for solo founders who do not want to read the matcher RFC.