SaaS gross margin benchmark: 2026
Best-in-class SaaS hits 80%+ gross margin. Here's what counts as COGS, why AI is compressing margins, and how investors use the number.
The short answer
Best-in-class SaaS hits 80%+ gross margin. The median for mature pure-play SaaS is around 75%. Below 70% is a structural warning sign that investors will discount your valuation multiple for.
The benchmark table
| SaaS type | Gross margin | Why |
|---|---|---|
| Pure B2B SaaS (Atlassian, Datadog) | 80–85% | Self-serve, low support |
| Mid-market SaaS | 72–80% | Some onboarding overhead |
| Enterprise SaaS (Salesforce, Workday) | 70–78% | Heavy implementation needs |
| SaaS + heavy services | 55–65% | Custom dev pulls down margin |
| AI-heavy SaaS (LLM inference) | 50–70% | Token costs eat into margin |
| Vertical SaaS (healthcare, legal) | 72–80% | Compliance overhead |
| Infrastructure / PaaS | 60–75% | Cloud cost pass-through |
| Marketplace-style (Shopify) | 50–60% | Payment processing cost |
What counts as COGS in SaaS
The biggest mistake founders make is mis-categorizing costs. SaaS COGS should include:
- Hosting and infrastructure. AWS, Azure, GCP, CDN, monitoring — typically 3–10% of revenue.
- Third-party APIs and data costs. Twilio, OpenAI, Stripe processing, data providers — 1–10%.
- Customer support. Tickets, live chat, phone for current customers — 5–10%.
- Customer success / onboarding. The portion directly tied to making customers successful — 3–8%.
- Professional services revenue COGS. If you sell implementation, the implementer salaries hit COGS, not OpEx.
SaaS COGS should not include: sales salaries, marketing spend, R&D / engineering for new features, finance, HR, or executive compensation. Those are operating expenses.
Why AI is changing the math
The big shift in 2025–2026 is that AI-heavy SaaS has structurally lower gross margins than traditional SaaS. A pure-software product serves one more customer for fractions of a cent. An LLM-powered feature can cost dollars per active user per month in inference fees. Companies like Notion, Intercom, and Salesforce are reporting margin compression of 3–8 points from AI features they bundled into existing plans.
The two ways to defend margin: (1) charge separately for AI-heavy features so unit economics stay healthy, or (2) drive inference costs down through smaller fine-tuned models, caching, and prompt engineering. Both are happening industry-wide.
The Rule of 40
The most-cited SaaS health metric is the Rule of 40: growth rate + profit margin should be at least 40%. A company growing 50% with -10% margin passes (50 - 10 = 40). A company growing 10% with 30% margin also passes (10 + 30 = 40). Failing the Rule of 40 typically results in a 25–40% valuation discount.
How investors use gross margin
Gross margin determines the ceiling on a SaaS company's eventual profitability. If gross margin is 50%, every dollar of revenue contributes 50¢ toward covering operating expenses. At 80%, it contributes 80¢. The difference compounds — at IPO scale, a 10-point gross margin difference often translates into a 2-3× revenue multiple difference.
Frequently asked
Is 60% gross margin bad for SaaS?
It's a yellow flag, not a red flag. AI-heavy, services-heavy, or infrastructure pass-through SaaS structurally runs lower. But for pure-play B2B SaaS, 60% means you have a cost problem to address before scaling.
Should customer success staff hit COGS or OpEx?
Split it. The portion serving existing accounts (renewals, support, expansion outreach) hits COGS. The portion working on new-business handoffs or top-of-funnel education is sales/marketing OpEx.
What's a good net margin for SaaS?
Mature, growth-moderating SaaS: 20–30% net. Growth-stage SaaS deliberately runs negative to fund expansion. Both can be healthy; it depends on your stage.