How Stripe Makes Money

The Toll Booth on Internet Commerce

Stripe started in 2010 with a famous simplification: seven lines of code to accept an online payment replacing the banking contracts and hardware that made online sales miserable. Fifteen years later the company processed about $1.9 trillion in payments in 2025 a 34 percent increase in one year and equivalent to about 1.6 percent of global GDP and a February 2026 employee tender valued it at159 billion dollars among the most valuable private companies in history not yet listed and still run by its founding brothers. The number in the headline invites the analyst question that this site continues to ask: of every dollar that flows what does Stripe really keep and for what

Anatomy of 2.9 Percent Plus 30 Cents

Stripe's classic sticker price for an online card payment in the US is 2.9 percent plus 30 cents. The critical fact of modeling is that most of it is not from Stripe. As the payment process article on this site describes each card transaction pays back to the customer's bank approximately 1.5 to 2.5 percent plus network assessments to Visa or Mastercard. Stripe which holds the position of processor of theacquirer collects the full rate remits those tolls and keeps the rest a net take rate which industry analysts estimate at around half a percent with which it funds its actual costs. So in honest terms Stripe's top line is rising at $1.9 trillion its true economics are a software margin on the slim slice it maintains and its operating leverage comes from the fact that processing the next $1 billion costs almost nothing incremental. The company said it was solidly profitable in 2025 notable in a sector where growth typically eats up margin

Working Through One Transaction

The step structure is easier to believe once you follow a single sale through it. Illustrative and round using the price tag and a mid-range exchange

A customer buys something for $100. Stripe collects 2.9 percent plus 30 cents which equals $2.90 plus 30 cents or $3.20

From there the exchange goes to the bank that issued the customer's card. At 2 percent that's 2 dollars. Network evaluations go to Visa or Mastercard a much lower figure call it 13 cents

Stripe takes about $107 on a $100 sale a little more than one percent and that pays for fraud systems engineering support and everything else

Now let's compare that to the estimated net acquisition rate of about half a percent across the business and see that they don't match up. That gap is the most informative thing about the model

The sticker price is what a small merchant pays. The 1.9 trillion is dominated by the big ones and the big merchants don't pay the sticker. They negotiate intensely quote on swap-plus terms where the pass-through is explicit and the processor's margin is a certain number and they move volume between suppliers to keep that number under pressure

Therefore the combined half percent is the weighted result of a small number of very large merchants paying very little and a long tail paying the published rate. Which means that the reported net take rate is as much a statement about customer mix as it is about pricing and will automatically decrease as Stripe gains larger accounts even if it never loses a single price negotiation

One more figure follows directly. Half a percentage point of 1.9 trillion is equivalent to about $9.5 billion of net income which is the figure that describes the business. The 1.9 trillion describes the pipeline

Why the 30 Cents Decides Who It Can Serve

The fixed component looks like a rounding detail next to the percentage and silently determines which companies can exist on these rails

Run the same sticker price on all ticket sizes and look at the effective fare

On a $100 sale the fee is $3.20 which is 3.2 percent. The 30 cents are barely noticeable

On a $20 sale the fee is 58 cents plus 30 cents or 88 cents which equals 4.4 percent

On a $5 sale the fee is 14 and a half cents plus 30 cents or about 44 and a half cents which is 8.9 percent

On a $1 sale the fee is about 33 cents which is 33 percent

The percentage stayed put the entire time. The fixed fee did all the work because a fixed fee is a larger part of a smaller number

That fact alone explains a lot about what the Internet sells. Micropayments never worked on cards not because no one wanted them but because the toll exceeded the price of the product. Companies that sell cheap digital goods bundle them into subscriptions sell credits in larger blocks or accept that a third of the income from small purchases disappears

It's also the specific gap that alt-rails continue to target and it's why Stripe's move into stablecoin infrastructure is more than a fad acquisition. A rail whose cost doesn't include a fixed floor per transaction changes what prices are viable and the current processor has good reason to own that option rather than watch someone else take advantage of it

The Second Act: Software on Top of the Flow

The strategic story of modern Stripe is bottom-up margin expansion. Payment processing itself is commoditized at the edges large merchants negotiate acceptance rate ruthlessly so Stripe sells software that ties to money in motion billing and subscription management invoicing tax calculation and remittances in thousands of jurisdictions fraud scoring in-person terminals and Connect the product powering marketplaces that split payments among millions of sellers. It also provides the pipeline for the wave of embedded finance thatcovers this site issuing cards and providing banking features within other companies' products and has planted its flag on stablecoin rails acquiring infrastructure to move money in the newly regulated tokens that our article on stablecoins describes. Management says that this revenue and finance automation suite alone is approaching a billion-dollar annual run rate small compared to the river of payments but with software margins and more importantly wrappingthe commodity tollbooth in switching costs. A merchant can ditch a processor. A merchant whose billing tax and payment logic resides in Stripe effectively cannot do so

What Connect Actually Solves

Connect gets a clause on that list and deserves more because it's the clearest example of how the switching cost argument works

Consider what a marketplace actually has to do when a buyer pays. A payment arrives that must be split between a seller the platform commission and sometimes a shipper or service provider. That's the easy part

The difficult part is everything around it. Every seller who receives money must be identified and verified according to the standards that regulators expect of anyone who moves funds on behalf of others. Someone has to be the merchant of record which determines who is responsible when a transaction is disputed. Payments must be made on schedule in the correct currencies to accounts in many countries. Tax filing must be done per seller. Disputes must be resolved when the counterparty is the seller and not the platform

Built in-house that's not a payments feature. It's a compliance and licensing program and it's why marketplaces historically took years to launch payments and then employed teams to keep them running

That's why Connect is sticky in a way that processing isn't. A merchant that only accepts card payments is buying something interchangeable and can get a better rate in a quarter. A marketplace running on Connect has its seller onboarding its identity verification its payment logic and its regulatory position around a single vendor. Moving is not an acquisition decision it's a restructuring project with a compliance review attached

That's the whole strategy in miniature. Sell the product at a competitive price and make sure it arrives wrapped in something that no one wants to rebuild

Why the Software Layer Carries Different Economics

It's worth being precise about why moving up the stack helps because both layers share the property that made the payoffs attractive in the first place

Processing the next billion dollars costs almost nothing incremental. Serving the next billing customer also costs almost nothing incremental. They are both near-zero marginal cost software so operating leverage is not the difference

The difference is pricing power and arises from comparability

A payment is a result of a commodity. Money arrives or it doesn't arrive and every serious processor delivers it. A merchant comparing suppliers is comparing one rate with another rate which is exactly the condition under which price falls toward cost. Larger merchants have the volume to continually make that comparison and they do

Billing logic tax determination in thousands of jurisdictions and fraud scoring are not comparable in that sense. They are built into the way the merchant's business works their output is difficult to compare and changing them means redesigning something that currently works. A merchant does not price tax remittances the same way it prices basis points

So the same marginal cost structure produces two very different outcomes. Identical economics on the way in and completely different ability to maintain the price on the way out

Which recasts the billion-dollar run rate. Against $9.5 billion in net revenue it seems like a rounded item and its importance is not its size. It's just that this is the part of the business where the price is set by Stripe instead of the merchant's next negotiation

How to Grade It Like an Analyst

Frame the bull and the bear honestly. The bull Stripe is a leveraged bet on the growth of Internet commerce with an take rate defended by software lock-in distribution across all startups formed in the last decade and optionality in AI-powered commerce where its agent payment tools aim to be the box when software starts buying software. Bearish net take rates compress the entire industry as merchants scale andRivals Adyen above all court the largest portfolios the pass-through structure means revenue growth favors the underlying economy and a $159 billion valuation already capitalizes on years of impeccable execution. The judgment comes down to one variable: whether the pricing power of the software layer outweighs the commoditization of the payments layer. That is also and not coincidentally the analytical template for every payments company you will ever value: separate the river from the splinter and thenask what defends the splinter

Making That Template Measurable

That template is only useful if it produces a number and it does which is worth explaining because it's the most useful habit for reading any payments business

Compare the growth rate of processed volume with the growth rate of net income

If volume grew by 34 percent and net revenue grew by 34 percent the acquisition rate held steady. If volume grew by 34 percent and net revenue grew considerably less the acquisition rate was compressed and the difference between the two rates is compression stated clearly

That comparison overcomes the reporting problem that identifies the bear case. Volume always looks impressive because it is the largest number available. Net income pays for everything and the gap between the two growth rates is precisely what is in dispute

The refinement is to ask why such a compression occurred because two very different stories produce it. Losing price negotiations is bad. Winning very large merchants who pay less per dollar is good as it increases absolute profit while lowering the average rate. A combined take rate falls in both cases and only one of them is a problem which is why mixed disclosure matters more than the overall rate

Then apply the same test to the software line and the entire thesis will resolve into one observable question. If net income growth is below volume growth while the software package is accruing faster than payments the second act is at work and compression is the cost of scale. If they both slow together the tollbooth is being commoditized and nothing is replacing it

The Bottom Line

Stripe earns a small slice of a huge and growing river $1.9 trillion processed by 2025 shifts most of its core fees to banks and networks and builds its future on software that makes it more expensive to exit now valued at $159 billion thanks to that combination. The company's true product was never payments but the elimination of financial complexity for anyone building on the Internet with a toll price. Judge and every imitator byThe same test rivers are rented chips are earned and only the software keeps them. And when you check compare net income growth with volume growth because that one comparison is where the argument really settles