Acquisitions

How to value a SaaS company when it runs on AI models

Knowing how to value a SaaS company is the starting point of every acquisition conversation. The method buyers use is simple to describe and harder to apply: pick the right earnings base, apply a multiple that fits the size and quality of the business, then adjust for risk. AI products add a few questions of their own, mostly about model costs and how defensible the product is. This guide explains the method step by step and frames every range as what it is, a general indication rather than a price.

Founder sketching revenue growth on graph paper beside a laptop

What the numbers in this guide mean

Every multiple in this guide is a general, indicative range drawn from public market commentary on small and mid-sized software deals. It does not describe any specific business, it is not an appraisal and it is not investment advice. Real prices depend on the buyer, the deal structure and facts that only diligence reveals. Acquisitions is a listing and deal-workflow platform, not a broker-dealer and not an investment adviser.

Choose the right earnings base first

Before any multiple, decide what you are multiplying. Online business deals generally use one of three bases.

  • Seller's discretionary earnings (SDE). Profit plus the owner's salary and personal expenses run through the business. Common for small, owner-operated businesses where the buyer will step into the founder's role.
  • EBITDA. Earnings before interest, taxes, depreciation and amortization. Used for larger businesses with a team and management that stay after the sale.
  • Annual recurring revenue (ARR). Used when the business is growing fast and reinvesting, so current profit understates its value, or when the buyer values strategic fit over cash flow.

Small AI SaaS products run by one or two founders are usually valued on SDE or on a mix of SDE and ARR. Venture-backed AI startups with fast growth and little profit are more often discussed in terms of ARR. Picking the wrong base is the most common reason founders and buyers talk past each other.

Normalize the numbers before you multiply

Buyers adjust reported numbers to reflect how the business will run under new ownership. Typical adjustments include adding back one-off costs, adding back the owner's salary for SDE, removing revenue that will not recur, and deducting the cost of replacing work the founder does for free. For an AI product, also check whether current model costs reflect promotional credits from a provider that will expire. If they do, normalized margin is lower than today's margin.

SaaS valuation multiples as indicative ranges

Public commentary from brokers and marketplaces on small software deals tends to place outcomes in broad bands. As a general orientation only:

Business profile Typical base Indicative range discussed publicly
Small owner-operated SaaS, stable, modest growth SDE roughly 2x to 5x annual SDE
Small SaaS with strong growth and low churn SDE or ARR upper end of SDE bands, sometimes priced on ARR
Larger SaaS with a team and steady growth EBITDA or ARR wider bands, often discussed as several times ARR
Micro SaaS with one founder and thin documentation SDE lower end of SDE bands
Pre-revenue product or domain only none priced on comparable asset sales, not revenue multiples

These bands are wide on purpose. Two businesses with the same revenue can sit at opposite ends because of growth, churn, margin and risk. Treat the table as a map of where conversations start, not as a quote. Our SaaS valuation multiples page explains the method our estimator uses and the factors it applies.

Factors that move an AI SaaS up or down a range

Once you have a base and a band, the position within the band depends on quality. Buyers look at the same handful of factors in almost every deal.

Growth

Year-over-year revenue growth is the single strongest driver. A business that grows steadily without heavy paid acquisition commands more than a flat one with the same revenue. Buyers also check whether growth is recent and accelerating or a single spike.

Churn and retention

Low monthly churn means revenue that compounds. High churn means the business has to replace a large share of customers every year just to stand still. We cover the maths in SaaS churn rate.

Gross margin after model costs

This is where AI products differ from classic SaaS. Classic software often has very high gross margins because serving one more customer costs little. An AI product pays an AI model provider for every request. If inference costs take a large share of revenue, the business is closer to a reseller than a software company and buyers price it lower. A product that controls costs through caching, smaller models or usage limits in its plans keeps more of each dollar.

Dependence on a single model provider

If the product works only with one provider's model and that provider changes prices, terms or capabilities, the business changes overnight. Buyers value products that can switch models without rewriting the core, and products whose value sits in workflow, data and distribution rather than in a single prompt.

Owner dependence

How many hours a week does the founder work, and on what? If the founder personally handles sales, support and engineering, the buyer is buying a job as well as a business. Documented processes and contractors who stay after the sale move the price up.

Customer concentration

If one customer brings a large share of revenue, losing that customer would change the business. Buyers usually discount for concentration or ask for an earnout tied to key accounts.

Age and track record

A business with several years of stable revenue is easier to underwrite than one that started recently. Short history usually means a lower multiple or a structure with deferred payments.

Quality of the numbers

Revenue verified at the payment processor and traffic verified at the analytics source reduce a buyer's risk. Self-reported spreadsheets do not. Verification does not change what a business earns, but it often changes how much a buyer is willing to pay for it, and how quickly. See Revenue verification for how read-only connections work on Acquisitions.

A worked calculation, step by step

Here is how the method fits together, using round illustrative numbers rather than any real business.

  • Reported annual revenue is 300,000 USD and reported annual profit is 120,000 USD.
  • The founder pays themself 60,000 USD through the business, so SDE before other adjustments is 180,000 USD.
  • Model costs are currently reduced by provider credits worth 20,000 USD a year that will expire. Normalized SDE is 160,000 USD.
  • Growth is steady and churn is low, but the founder handles all support. The buyer places the business in the middle of a general SDE band.
  • Applying a mid-band multiple gives a range rather than a point. If the middle of the band is around 3x to 4x, the indicative range is 480,000 to 640,000 USD.
  • Structure then shapes the final number. A buyer may offer the lower end in cash at closing, or the upper end with part deferred or tied to revenue retention.

The output is a range for discussion, not a valuation of record. Different buyers will land at different points within it.

Questions buyers ask that are unique to AI products

  • What share of revenue goes to inference and hosting, and how has that share changed over the last year
  • What happens to quality and cost if you switch to another model
  • Do you own the prompts, evaluation sets and any fine-tuning data, and can they be transferred
  • How do you prevent heavy users from costing more than they pay
  • Are there usage limits in each plan, and are they enforced in code
  • What does the product do that a general-purpose AI assistant cannot do on its own

Clear answers to these questions often matter more than an extra point of growth. A seller who has them ready shortens diligence and supports a price toward the top of the range. Our guide on SaaS metrics lists the numbers to prepare.

Common valuation mistakes

  • Applying a headline multiple from public company news to a small private business. Public company multiples reflect scale, liquidity and growth that small deals rarely have.
  • Multiplying revenue when the buyer is thinking in profit, or the reverse.
  • Ignoring deferred revenue from annual plans.
  • Treating promotional model credits as a permanent cost advantage.
  • Anchoring on a single number instead of a range.

Getting an indicative range on Acquisitions

The valuation estimator on Acquisitions applies the method in this guide: it picks a base from your inputs, places the business in a general band for its category and adjusts for growth, churn, margin after model and hosting costs, size, age and owner hours. The result is always a range with the factors that moved it, labelled as an indicative estimate and not an appraisal. Sellers on every plan see a valuation range on their listing, and buyers on the Scout, Acquirer and Fund plans run estimates on any business they review.

FAQ

Fast-growing businesses that reinvest everything are often discussed on a multiple of ARR rather than profit. Buyers then weigh growth, retention and the path to profitability. The range is usually wider and more dependent on the buyer.

Not automatically. Strong growth and demand can support higher interest, but higher model costs and provider dependence can pull margins and multiples down. Each business is judged on its own numbers.

No. It is an indicative estimate based on general public market ranges and the inputs you provide. It is not an appraisal, a fairness opinion or investment advice.

Most micro SaaS businesses run by one founder are discussed on SDE, often toward the lower part of general bands because of owner dependence and short history. Good documentation and low churn help move them up.

No. A domain without a product has no revenue to multiply. It is priced on comparable domain sales and brandability, which a revenue model does not capture.

See where your business sits in the range

Run the estimator on SaaS valuation multiples, compare with Verified AI business listings, or look at Micro SaaS marketplace if your product is run by one or two founders.