AS | Ankit Sarawagi|Founder, CFOmatrix·July 2026·10 min read | Free template |
- The structure is fixed: revenue, less cost of services, equals gross profit and gross margin, less operating expenses, equals EBITDA.
- COGS is delivery, opex is building and selling. The two classic mistakes are support in opex and research and development in COGS. Both distort the one number investors trust most.
- AI changes the margin line: variable compute pulls gross margin from about 80 percent to around 65 percent, and you must read it per customer, not blended.
- EBITDA is the profit line boards track, read alongside gross margin and growth, not in isolation. Grab the free template below and follow the same layout every month.
| 2 lines Gross margin and EBITDA carry most of investor trust | 80% → 65% Gross margin: classic SaaS vs AI SaaS today | 2 mistakes Support in opex, R&D in COGS: both break the margin |
Why the SaaS P&L Gets Read Differently
Every business has a profit and loss statement, but investors read a SaaS one in a particular order. They do not start at net profit. They start at gross margin, because that tells them whether the unit economics work, then they look at the EBITDA line to see how far the whole business is from paying for itself. Everything in between (how you split revenue, what you call cost of services, where you park research and development) either supports that story or quietly undermines it. A P&L built on the standard SaaS structure is trusted at a glance; an idiosyncratic one invites questions you do not want in a diligence room. This post is part of our SaaS finance pillar guide, focused on the one statement founders most often build wrong.
The SaaS P&L, Line by Line
Here is the whole statement in the order it should appear, from revenue at the top to EBITDA at the bottom. Build yours in exactly this shape and an investor can follow it without a legend.
Notice that the entire statement hinges on one boundary: the line between cost of services (above gross profit) and operating expenses (below it). Put a cost on the wrong side of that line and both your headline margins move. That boundary is worth its own section.
COGS vs Opex: What Goes Where
The rule is simple to state and easy to get wrong: COGS is the cost of delivering the product to a paying customer; opex is the cost of building and selling it. If a cost stops the live service the moment you cut it, it is COGS. If cutting it only slows future growth, it is opex.
| Cost | Where it goes | Why |
| Cloud, compute, hosting | COGS | Directly runs the live service |
| Third-party model APIs | COGS | Consumed per request to deliver output |
| Delivery support & customer success | COGS | Delivers the service to live accounts |
| Payment processing fees | COGS | Incurred on every rupee collected |
| Sales & marketing | Opex | Wins future customers, not delivery |
| Research & development (new product) | Opex | Builds tomorrow’s product, not today’s service |
| General & administrative | Opex | Runs the company, not the product |
Founders flatter gross margin by pushing delivery support into opex (so it disappears from cost of services), and they accidentally sink gross margin by dumping research and development into COGS. Both are red flags in diligence. Support that keeps live customers running is COGS; engineers building next quarter’s features are opex.
The Gross Margin Line Investors Read
Once COGS is honest, gross margin falls out on its own. Classic SaaS lands at 80 percent plus because the marginal cost of one more customer is close to zero. AI SaaS today lands at around 65 percent plus, because a real, variable compute bill sits inside cost of services. The bridge below shows where that 15 points goes.
Investors do not penalise a lower AI margin, they penalise a fake one. A clean 65 percent with every delivery cost in COGS earns far more trust than an 82 percent that quietly leaves support and inference out. Show the honest number and show it rising. More depth in SaaS gross margin and cloud COGS.
The AI SaaS Twist: Segregate R&D, Read Per Customer
AI SaaS does not just lower the margin, it changes how you have to build the P&L. Two disciplines matter, and both come from running product and production at the same time.
“A product company builds tools and runs production simultaneously, so you have to segregate research and development cost from cost of services. If you do not, your gross margin is fiction. And in AI SaaS you cannot read margin blended any more, you read it per enterprise customer, per geography, because compute varies so much.”
Ankit Sarawagi, from working across AI SaaS startupsThe first discipline is the research and development split. The same engineering team often builds next quarter’s features (R&D, opex) and keeps the live model serving customers (cost of services, COGS). Split their cost deliberately, usually by time or headcount allocation, or your gross margin is meaningless. The second is the per-customer view. Because one customer can burn ten times the compute of another on the same plan, a single blended margin hides who is profitable and who is underwater.
Add metering and per-customer cost tags to your compute the moment you launch, not after you have a hundred accounts. It is not complicated to do early and painful to retrofit late. It is what lets you decide which accounts to reprice at renewal. See the hidden costs of SaaS for the lines founders miss.
⬇ Free download The SaaS & AI SaaS P&L template A ready-to-use layout: revenue split, cost of services, gross margin, opex and EBITDA, with the COGS-vs-opex classification built in and a per-customer margin tab for AI SaaS. Drop in your numbers. | Get the template |
EBITDA: The Right Profit Line, Read the Right Way
Below opex sits EBITDA, earnings before interest, tax, depreciation and amortisation. It is the operating profit line SaaS boards and investors track, because it strips out financing and accounting effects and shows whether the core business makes money. The nuance: early-stage SaaS is usually EBITDA negative on purpose, spending on growth ahead of profit. So it is never read alone.
Common SaaS P&L Mistakes to Avoid
“A P&L an investor trusts is not the one with the highest margin. It is the one where every cost is on the side of the line it belongs on, and the number still holds up when they push on it.”
Ankit Sarawagi, CFOmatrixFrequently Asked Questions
How do I structure a SaaS P&L?
Top to bottom: revenue (split into subscription, usage and services), less cost of goods sold or cost of services (compute, model APIs, support and customer success, third-party delivery, payment processing), which gives gross profit and gross margin percent; then operating expenses split into sales and marketing, research and development, and general and administrative; which gives EBITDA. The two lines investors read first are gross margin and EBITDA, so keep the COGS-versus-opex split clean and consistent every month.
What goes in COGS versus opex for a SaaS company?
COGS (cost of services) is everything directly needed to deliver the product to a paying customer: cloud and compute, third-party model APIs, databases and pipelines, delivery support and customer success, other third-party delivery providers, and payment processing. Operating expenses are the costs of building and growing the business: sales and marketing, research and development, and general and administrative. The classic mistakes are hiding delivery support in opex and dumping research and development into COGS.
What gross margin should a SaaS company show investors?
Traditional SaaS runs at about 80 percent plus. AI SaaS today runs at around 65 percent plus, because compute is expensive and variable and the market is not yet paying a premium for AI value. Investors accept the lower AI margin as long as COGS is honest and you can show it rising over time. The number that damages trust is not a lower margin, it is a margin propped up by leaving delivery costs out of cost of services.
Where does R&D go in a SaaS P&L?
Research and development belongs in operating expenses, below gross profit, not in COGS. This matters most for product companies that build new tools and run production at the same time: engineers building next quarter’s features are R&D (opex), while the compute and people running the live service are cost of services (COGS). Fold R&D into COGS and gross margin looks worse than it is; fold delivery into R&D and it looks better. Segregate the two deliberately.
How is an AI SaaS P&L different from a traditional SaaS P&L?
Two things change. COGS carries a large, variable compute line (own models, third-party model APIs, databases, pipelines) that moves with usage, so gross margin sits around 65 percent rather than 80 percent. And because compute varies by customer, a single blended margin hides the truth: you read margin per customer and per geography, which means metering and per-customer cost tags built in from day one. Product AI companies must also segregate research and development from cost of services carefully, because they build and run at the same time.
Is EBITDA the right profit line for a SaaS company?
EBITDA (earnings before interest, tax, depreciation and amortisation) is the operating profit line most SaaS boards and investors track, because it strips out financing and accounting effects and shows whether the core business makes money. Early stage, most SaaS companies are deliberately EBITDA negative while they invest in growth, so it is read alongside gross margin, growth rate and the Rule of 40, not on its own. Gross margin tells you if the unit economics work; EBITDA tells you how far the whole business is from self-funding.
This is general educational information for founders, current to mid-2026, drawing on the author’s experience across SaaS and AI SaaS startups, and is not legal, tax or investment advice. Margin benchmarks are indicative and vary by stage and business model. Accounting classification (revenue recognition, cost allocation) should follow the applicable standards; verify the current position or consult a professional before acting on a specific matter.
The Hidden Costs of SaaS
Deferred Revenue and Billing
→ Back to the SaaS Finance pillar guide
AS | Founder, CFOmatrix | Finance Strategy & Equity Compliance CFOmatrix is a knowledge platform focused on how finance actually works inside growing companies. This guide draws on hands-on experience across SaaS and AI SaaS startups, from building the P&L and reading per-customer margin to pricing, structure and investor readiness. |