How much do AI startups make? Reading verified revenue data

Most AI startups make very little money, a small group makes a startling amount, and the gap between the two is the story. The only way to know which group a company belongs to is to look at revenue that was verified, not revenue that was claimed on social media or guessed from traffic.

This guide explains what "verified" means, what recent reports say about typical and exceptional AI companies, and how to read any revenue figure without being fooled.

Why headline numbers mislead

Revenue stories travel fast because the extreme cases are the interesting ones. A product that reaches eight figures in a year gets a thousand posts. A product that earns $400 a month gets none.

That creates survivorship bias. If you only see the winners, the average looks far better than it is. Treat every public number as a sample from the top of the distribution, not from the middle.

There is a second problem with definitions. "Revenue" can mean cash collected, recognized revenue, monthly recurring revenue, or annualized run rate. A founder who multiplies one great month by twelve and calls it ARR is not lying exactly, but the figure may not describe a stable business. We cover these terms in our plain-English guide to MRR and ARR.

What verified revenue actually means

Verified revenue is a figure that someone checked against a system of record. For a subscription product, that usually means the payment provider, such as Stripe, rather than a screenshot or a spreadsheet.

The levels of trust look like this:

Source of the number	How much to trust it
Founder tweet or screenshot	Low, easy to edit or cherry-pick
Founder interview with details	Medium, but still self-reported
Connected payment provider data	High, read directly from payments
Audited financial statements	Highest, but rare for small startups

The Buy a startup marketplace on StartupsIntel lists businesses with revenue verified from the payment provider, which is the standard a buyer should expect. If a seller cannot connect a payment account, treat the number as a claim.

What the big reports say

Benchmarks from investors and data providers give a useful frame, with the caveat that they cover companies large enough to be tracked.

Bessemer Venture Partners published The State of AI 2025, which split fast-growing AI companies into two groups. Its "Supernova" group averaged about $40 million in ARR in the first year and about $125 million in the second, but with gross margins around 25 percent. Its "Shooting Star" group averaged about $3 million in year one and about $12 million in year two, with margins near 60 percent.

Those are extraordinary outcomes. They describe companies that were selected because they grew fast, so they are a ceiling to admire and not a forecast to plan around.

Standard Metrics, which analyzes data from private companies, reported in its 2025 year in review that AI and non-AI companies with more than $20 million in annualized revenue grew at similar rates overall, around 30 percent year over year for AI by the fourth quarter of 2025. The top tenth of AI companies grew much faster, at 262 percent. The lesson is that the average is ordinary and the tail is enormous.

What a typical small AI startup earns

Public benchmarks rarely cover the long tail of small products, so be honest about what is known and what is not. We do not have a reliable census of tiny AI tools, and nobody else does either.

What can be said with confidence is how revenue behaves at the low end:

• Many AI tools are thin layers on top of a model provider, so they are easy to copy and hard to defend.
• Free tiers are common, so most users never pay.
• Traffic often comes from a launch spike that fades within weeks.
• Revenue is concentrated in a small share of paying accounts.

If you want a rough sense of where a specific product sits, look at its traffic and pricing, then apply a conversion assumption. Our method multiplies monthly visits by a 3 percent paying share and the lowest paid plan, split 80 percent monthly and 20 percent yearly. That is a labelled estimate, not a measurement, and the methodology page shows the assumptions.

How to read a revenue figure in five checks

Before you believe any number, run it through these questions.

1. What is the time frame? One month, a trailing twelve months, or an annualized run rate are very different things.
2. Is it recurring? One-time sales, consulting income and credit top-ups do not behave like subscriptions.
3. Is it gross or net? Refunds, chargebacks, app store fees and taxes can shrink the real figure.
4. Who verified it? Prefer a payment provider connection over a screenshot.
5. What are the costs? Revenue without margin says little. AI products carry inference costs that classic software does not. We unpack that in unit economics for AI products.

A product with $30,000 a month in revenue and a 25 percent gross margin has a very different future from one with the same revenue and a 70 percent margin.

Using traffic as a sanity check

Revenue should roughly track audience. If a site claims $200,000 a month but its traffic checker report shows a few thousand visits, something does not add up, unless the product sells to a small number of high-priced enterprise buyers.

Traffic is not a substitute for verification. It is a quick plausibility test. We explain how to turn visits into a rough revenue range in how to estimate a SaaS company's revenue from its traffic.

The monthly revenue rankings apply the same method across many AI products, always labelled as estimates. Use them to compare products with each other, not to read an exact bank balance.

Red flags in revenue claims

Watch for these patterns when a founder, a seller or a newsletter shares a number:

• A single best month presented as the normal month.
• Revenue "run rate" based on a launch week.
• No mention of refunds or churn.
• Screenshots that crop out the date or the currency.
• Growth percentages without the starting point. Going from $100 to $400 is 300 percent growth and still tiny.
• Claims that cannot be tied to any named payment account.

None of these proves bad faith. They simply mean the number has not earned your trust yet.

A practical way to use this

If you are an investor or buyer, ask for read-only payment provider access, then compare it with traffic and category norms. If you are a founder, publish verified numbers because it builds more credibility than any screenshot. If you are a marketer or researcher, label estimates as estimates and quote ranges rather than single points.

The honest answer to "how much do AI startups make" is a wide distribution. A few make millions, a larger group makes enough to cover costs, and a very large group makes close to nothing. Verified data is how you tell them apart.

Revenue is only part of the picture

Even a verified number needs context before it tells you anything about quality. Two products with the same monthly revenue can be worlds apart.

Ask how old the business is, because $20,000 a month after two years means something different from $20,000 after two months. Ask how many customers produce it, since a hundred small subscribers are safer than three large ones. Ask how much of the revenue comes from the founder's own audience, which may not carry over to a new owner.

Finally, ask what happens if the underlying model provider changes its prices or terms. AI startups depend on suppliers more than most software companies do, and that dependency shows up in margins long before it shows up in revenue. A business that earns steadily and keeps most of what it earns is worth far more than one that earns loudly and keeps little.

FAQ

How much does the average AI startup make?

There is no reliable average for small AI products because most do not publish figures. Reports from Bessemer and Standard Metrics describe companies that are already large or fast growing, so they overstate what a typical new product earns. Plan around a modest outcome and treat large numbers as upside.

Is revenue shown in rankings real?

Rankings on StartupsIntel are modelled estimates from traffic and a pricing assumption, and they are always labelled that way. Real, verified revenue comes from a connected payment provider, which is the standard used for listings on the marketplace.

What is the difference between revenue and profit for an AI startup?

Revenue is what customers pay. Profit is what remains after costs, and for AI products those costs include model inference, which can take a large share of every dollar. Always ask about gross margin alongside revenue.

Can I trust a founder's screenshot of Stripe?

A screenshot is a starting point, not proof. It can be edited, cropped or taken from a best month. Ask for read-only access to the payment dashboard or a verified listing before you rely on the number.