2026 SaaS Gross Margin Report: The Median Is 80% on Software, Not 75%

Boards quote 75 percent. That is the investor floor Aleph still prints for software delivery, not a 2026 median. The typical private B2B SaaS company is at 80 percent on software and 76 percent once services are in. Usage-only pricing sits at 62 percent. AI was supposed to crush the number. At the median, it has not.

Gross margin is the first dollar of the P&L: what is left after you deliver the product, before you fund research, sales, marketing, and profit. It is not the Rule of 40. It is not ARR per employee. It is the ceiling those scores have to live under. This B2Bcentr report separates software from blended, subscription from usage, and the 75 percent bar from the 80 percent median.

Key Takeaways

• The median private B2B SaaS software gross margin is 80% on full-year 2025 actuals (232 of 342 companies in the 1 June 2026 Aleph × Benchmarkit report). The median total-revenue gross margin is 76% (228 companies). Top quartile: 86%+ on software, 84%+ on total.
• Boards still quote 75–80%+ on software as what investors look for. That is the floor. Typical in this sample is already at the top of that band on software, and at the low end of Aleph’s 75–85% “healthy” range once services are blended in.
• Usage-only pricing posts a 62% median total gross margin, against 76–84% for subscription variants. The same report: usage-based companies lead on net revenue retention (108%) and ARR per employee ($291,000). The margin is the trade-off, not a miss.
• Software margin held at 79–81% across four years. Total-revenue margin slipped from 77% to 76%. The 25th percentile on software improved from 70% to 72%. The floor of the bottom quartile is 50% — LLM inference, infrastructure, and per-customer customization. Do not average 50 and 72.
• Scale is a 14-point lever: 72% software margin below $5 million ARR, 86% at $50 million–$100 million (band 84–90%). Companies growing more than 50% sit at 74% total margin. The 21–30% growth band is the high print, at 78%.
• 67% of companies that already have AI in the product do not charge extra for it (156 respondents). 29% charge separately. The median software margin held anyway. Instrument inference now; next year’s benchmark will split AI-native product margins.
• SaaS Capital’s March 2026 survey of more than 1,000 private B2B SaaS companies does not publish a 2026 gross-margin median. It publishes the COGS lines: hosting 5% of ARR, DevOps 4%, professional-services COGS 5%, other COGS 3%. Those are the dollars the 80 / 76 split is dividing. Adding those medians is not a surveyed margin.

Three samples, three different margins

Search “what is a good SaaS gross margin” and you will get 75 percent, or 80, or 70. The useful question is which sample produced the number sitting next to it.

The investor floor, still 75–80%+ on software. Aleph’s 2026 page is explicit: investors typically look for 75–80% or higher on software, because that is what funds R&D, sales and marketing, and profit. A print below about 70% usually signals a delivery-cost or services-mix issue. This is the bar, not a 2026 median. SaaSDB’s public-company guide (200-plus SEC filers, January 2026) rates 80%+ best-in-class and 70–80% strong. It does not print a single live public median in HTML — only “approximately 72–76%” — so this report does not invent one.

Aleph × Benchmarkit, 1 June 2026. 342 B2B SaaS and AI-native software companies; software gross margin from the 232 that reported it, total-revenue gross margin from 228. Full-year 2025 actuals. Median 80% software, 76% total. Top quartile 86%+ / 84%+. This is the private-operator number for 2026 planning. Aleph’s own page is explicit: the report is a 2026 edition of 2025 actuals.

SaaS Capital, 10 June 2026. 15th annual survey, completed March 2026, more than 1,000 private B2B SaaS companies — the same sample behind B2Bcentr’s marketing budget report. No 2026 SaaS Capital gross-margin median is published on the page fetched, so none is invented here. What the survey does print is the cost stack: hosting 5% of ARR, DevOps 4%, professional-services COGS 5%, other COGS 3%, customer support and success 9%. Those lines are the denominator. They are not one blended margin.

If your board is using 75 and your finance team is using 80, you are not disagreeing about health. You are disagreeing about software versus total, and about the bar versus the median. If someone is using 62, they are looking at usage-only. Put the cut on the slide.

Software versus total is the formula argument

The formula is simple:

Gross margin = (Revenue − cost of goods sold) ÷ Revenue × 100

Two conventions change the answer more than the decimal.

Lock software or total. Software gross margin covers subscription revenue against its direct delivery costs — hosting, infrastructure, support, third-party software, and, increasingly, AI inference. 2025 median: 80%. Total-revenue gross margin includes services, onboarding, and other non-recurring revenue. 2025 median: 76%. The 4-point gap is the services drag. A wide gap is a services-heavy model, not a hosting problem.

Lock what sits in COGS. Hosting and cloud. Customer support. Third-party APIs. Payment processing. Professional services, if they are material. R&D, sales, marketing, and G&A are below the line. Misclassifying customer success into sales and marketing inflates reported gross margin. Write the definition at the top of the slide. Publish the two inputs, not just the percentage: revenue this period, COGS this period, software and total on separate rows.

An 80 percent software print means 80 cents of every software dollar is left to fund the rest of the company, at the median firm that reported it. It does not mean the Rule of 40 is 80. Rule of 40 is growth plus operating profit. Gross margin is the ceiling that operating profit lives under. The Magic Number divides sales and marketing, not COGS. Do not convert one into another.

AI did not dent the median. It is already in COGS.

Median software gross margin held inside 79–81% for four years — CY-2022 through CY-2025. Industry-wide LLM and AI infrastructure costs have not compressed software margin at the median. Total-revenue margin slipped one point, 77 to 76, which the report reads as transitional cost from integrating AI into existing products, not a structural break. The 75th percentile on total margin improved, from 80% in CY-2022/23 to 84% in CY-2024/25. The 25th percentile on software improved from 70% to 72%.

The bottom of the distribution is a different company. Benchmarkit’s software tables put the first-quartile floor at 50% — firms not yet optimizing delivery, where inference, cloud, and per-customer customization pile up. The 25th-percentile cutoff is 72%. Those are not the same cut. 50 is the worst of the bottom quartile. 72 is where the bottom quartile ends. Do not average them.

The AI bill is already a product decision. Of 254 companies in the AI-maturity cut, 35% first shipped AI in the product in the past 12 months. Of the 156 that already have AI functionality and answered the monetization question, 67% do not charge extra for it. 29% charge separately. Only 13% of those who do charge use usage-only pricing for AI; subscription-plus-usage is the common hybrid at 39%. The report could not prove an AI-compression hypothesis with the data it had. Next year’s benchmark will isolate AI-native product margins. Instrument inference, RAG, and model cost as their own COGS lines now. A company absorbing GPU spend as a free feature is running an experiment on that 80.

Pricing model and scale are the filters

Copying 80 percent onto a usage-only product, or 62 percent onto a seat-based mid-market motion, is how the plan breaks.

Usage-only is the structural exception. Consumption revenue carries higher infrastructure and compute, especially before unit COGS normalize. The same companies post the strongest NRR (108%) and the highest ARR per employee ($291,000). You are trading margin for expansion that does not need a new closer for every dollar. Non-seat subscription — often priced on a fixed minimum commitment — leads the subscription cluster on margin. Hybrid fixed-fee plus usage outperforms on software margin in the Benchmarkit tables, likely because the usage overlay is cost-plus.

Software margin rises 14 points from sub-$5 million ARR (72%) to $50–$100 million (86%). At that band the distribution is tight, 84–90%: execution, not structure, decides where you land. Below $5 million, 72% is a normal early-stage print, not a failed SaaS model. The 21–30% growth cohort has the highest total margin (78%). Companies growing more than 50% sit at 74%, accepting near-term services and CS load to take share. High growth and high gross margin can coexist. They are not the same slide as a 2.40 Magic Number.

The dollars the percentage is dividing

A gross margin without a COGS stack is a slogan. SaaS Capital’s 2026 department medians are the private-B2B cost lines, not a surveyed gross margin. Convert the COGS rows at $5 million ARR. These are arithmetic from published percentages, not a surveyed combined COGS figure.

At Aleph’s 80% software median, $5 million of software revenue leaves $1 million of software COGS and $4 million of software gross profit — arithmetic from the published ratio, only if your company is in that 232-company cut and the revenue is actually software. At 76% total, the same $5 million of mixed revenue leaves $1.2 million of blended COGS. The 4-point gap is $200,000 at that scale. That is the services drag in dollars.

The $3 million–$5 million ARR band inside SaaS Capital is a different mix: hosting still 5%, DevOps 3%, professional-services COGS 5%, other COGS 3.5%, customer success 10%, selling 12%, marketing still 8%. If your “gross margin problem” is actually success headcount coded as COGS, you have a classification fight, not an AWS fight. If hosting is already 5% and software margin is 72%, you are in the sub-$5 million band, not the $50 million band.

Customer success at 9% of ARR sits on the border. Some operators put a slice of it in COGS (the support that delivers the product). Some put it in opex (the team that renews and expands). Mixing those treatments is how two companies at the same AWS bill print 80 and 76. Lock the coding before you lock the target.

Gross margin is also the input the rest of the 2026 series uses without saying so. Lifetime value in a CLTV:CAC ratio has to be gross-margin-adjusted, or the 4.1x Aleph median is flattered. The margin half of Rule of 40 cannot exceed what gross margin allows after opex. A leaky base is already sized in the churn report: 84% gross revenue retention means you are replacing 16% of the book before expansion does any work. High gross margin with 84% GRR is a profitable leak. It is not a healthy one.

How to set your number without copying 75

1. Pick the sample that matches you. Private B2B software delivery: Aleph’s 80% median, 86%+ top quartile, 72% 25th percentile, 50% floor, on 2025 actuals. Private blended: 76% median, 84%+ top quartile. Usage-only: 62% total as the cohort, not as a miss against 80. Investor bar: 75–80%+ on software. Public comps: use a named 10-K, not SaaSDB’s approximate 72–76% range.
2. Lock software and total, and lock COGS. Two rows, every period. Hosting, support, third-party, inference, and services called out. Do not move customer success across the line when the print looks better.
3. Read it by stage and pricing model, not the headline. Target ~86% software at $50–$100 million ARR. Judge sub-$5 million on trajectory toward 72, then 80. A usage-based company at 65% is near its cohort. A seat-based subscription company at 65% has a delivery-cost problem.
4. Pair it with GRR, Magic Number, and the spend mix. When Magic Number, CAC payback, NRR, and GRR are all top-quartile, accelerating GTM investment lifts growth and the Rule of 40. Gross margin is the fuel those investments burn. If software is 80 and total is 68, you have a services-mix problem. If selling is already 15% of ARR and marketing is residual, another campaign will not fix COGS.
5. Instrument AI as its own COGS line. 67% of AI-in-product companies are not charging extra. That can be a land-and-expand choice. It is also how a four-year 80% median becomes next year’s surprise.

A one-page math check: software and total margins, services called out; 75 / 80 / 76 / 86 / 62 / 72 / 50 circled; hosting, DevOps, professional-services COGS, and other COGS next to SaaS Capital’s 5 / 4 / 5 / 3; inference as its own line. If those numbers do not fit on one slide, the 75 was never your number. Change the sample, the cut, or the stage assumption. Do not change the investor floor.

FAQ

What is a good SaaS gross margin in 2026?

Investors still look for 75–80% or higher on software, with 86%+ best-in-class in Aleph’s 2025 actuals. The private median is 80% software and 76% total. Usage-only sits at 62%. “Typical” on software is already the bar. The 50% floor is still a problem.

What is the difference between software gross margin and total gross margin?

Software gross margin covers only subscription revenue and its direct delivery costs (80% median). Total gross margin includes lower-margin services and other revenue (76% median). The 4-point gap is the services drag. For unit economics, software is the cleaner number. Total tells you how much implementation is doing to the blend.

How do you calculate SaaS gross margin?

Subtract cost of goods sold from revenue and divide by revenue. For SaaS, COGS includes hosting, support, third-party software, and increasingly AI inference — which are best tracked separately. Do not put R&D, sales, or marketing in COGS. Do not switch customer success from COGS to opex between periods.

Has AI reduced SaaS gross margins?

Not at the median. Software gross margin held at 79–81% across four years. Total-revenue margin slipped from 77% to 76%. The report could not confirm AI-driven compression with available data and will isolate AI-native product margins next year. 67% of companies with AI in the product are not charging extra for it. Instrument the cost anyway.

Why is usage-based SaaS gross margin lower?

Usage-only models post a 62% median total gross margin because consumption revenue carries higher infrastructure and compute, especially before unit COGS normalize. Subscription variants cluster at 76–84%. The trade-off in the same survey is 108% NRR and $291,000 ARR per employee. A usage company at 65% is near cohort. A subscription company at 65% is not.

Does SaaS gross margin improve with scale?

Yes, on software. The median rises from 72% below $5 million ARR to 86% at $50–$100 million — a 14-point expansion through operational leverage. At that scale the band is 84–90%. Below $5 million, judge the trajectory, not the 80 percent headline.

Methodology

Figures are taken from primary publications, not from secondary “SaaS gross margin 2026” roundups.

• Aleph × Benchmarkit, What’s a good SaaS gross margin? (2026 benchmarks) and the 2026 SaaS and AI Metrics Benchmarks (1 June 2026). 342 companies; software GM n=232; total GM n=228; AI monetization n=156; AI-in-product timing n=254. CY-2025 actuals.
• SaaS Capital, 2026 Spending Benchmarks for Private B2B SaaS Companies, 10 June 2026. 15th annual survey, completed March 2026, 1,000-plus private B2B SaaS companies. Hosting 5%, DevOps 4%, professional-services COGS 5%, other COGS 3%. No 2026 gross-margin median is published on the page fetched, so none is invented here.
• SaaSDB’s January 2026 public-company guide was fetched for rating bands only. The “approximately 72–76%” public median is not used as a point estimate.

Dollar examples at $5 million ARR are arithmetic from published percentages and the 80 / 76 medians. Related Aleph medians cited as context (Magic Number 1.37, ARR per employee $291,000 usage-based, NRR 108% usage, GRR 84%) are from the same 1 June 2026 report.