How Stripe Makes Money: The Payments Take Rate

Stripe collects $3.20 on a $100 online card payment. Almost none of that is revenue.

Understanding how Stripe makes money starts with separating what the company collects from what it keeps. Stripe’s published US rate is 2.9% plus 30 cents per successful transaction (Stripe pricing page, accessed July 2026). That is a collection number. Interchange goes to the bank that issued the card, assessments go to the card network, and the processor keeps only the residual spread.

Research firm Sacra estimates Stripe converts roughly 3% in gross fees into about a 0.40% net take rate after interchange, network and partner costs (Sacra, Stripe company research page, accessed July 2026). On that estimate, close to seven of every eight dollars Stripe collects belongs to somebody else.

A spread that thin only becomes a large business at extraordinary volume, and it only holds up if the company can sell something else on the same flow. Both halves of that sentence are the business model.

One caveat governs everything below. Stripe is private and files no Form 10-K, so every Stripe figure here is company-announced or a named third-party estimate, never an audited disclosure. The public comparables carry the verification load.

Key takeaways

• The headline rate is a collection number. Stripe’s published US online rate is 2.9% plus 30 cents, which is $3.20 on a $100 charge (Stripe pricing page, July 2026). Sacra estimates only about 0.40% of volume survives as net take rate.
• Interchange is the layer that eats the fee. Stripe’s own documentation names three components: interchange to the issuing bank, assessments to the card network, and the processor markup (Stripe, Interchange-plus pricing explained, January 2026). Federal Reserve data puts debit interchange alone at 0.45% of value for covered issuers and 1.41% for exempt issuers (Regulation II biennial report, data year 2023).
• Scale is what makes a thin spread work. Stripe company-announced $1.9 trillion of total volume in 2025, up 34%, or roughly 1.6% of global GDP (Stripe 2025 annual letter, February 24, 2026).
• The public benchmark is thinner still. Adyen’s FY2025 net revenue of 2,364.2 million euros on 1,394.3 billion euros of processed volume implies roughly 17 basis points, at a 53% EBITDA margin (Adyen H2 2025 results, February 2026).
• Software on top of the flow is the margin story. Billing at 0.7% of billing volume, Connect at a 0.25% starting fee for platform-controlled pricing, and Radar for Fraud Teams at 2 cents per screened transaction all price against volume Stripe already carries (Stripe pricing page, July 2026).
• The disclosure gap is real. Stripe’s most recent tender valued it at $159 billion (company-announced, February 2026), while Sacra reports a January 2026 internal 409A mark of $106.7 billion. No audited statement reconciles the two.

---

How much does Stripe charge per transaction?

Stripe’s published US rate for online card payments is 2.9% plus 30 cents per successful transaction, with 0.5% added for manually entered cards, 1.5% for international cards, and 1% when currency conversion is required (Stripe pricing page, accessed July 2026). In-person payments through Stripe Terminal are priced at 2.7% plus 5 cents.

The percentage gets the attention. The fixed 30 cents does the damage.

Because the flat component does not scale with ticket size, the effective rate a merchant actually pays is a curve, not a number. Running Stripe’s published rate across ticket sizes shows how steep that curve is at the small end.

Charge amount	Fee at 2.9% + $0.30	Effective rate
$10.00	$0.59	5.90%
$20.00	$0.88	4.40%
$50.00	$1.75	3.50%
$100.00	$3.20	3.20%
$250.00	$7.55	3.02%
$500.00	$14.80	2.96%
$1,000.00	$29.30	2.93%
$5,000.00	$145.30	2.91%

Author calculation applying Stripe’s published US online card pricing (2.9% + 30 cents, Stripe pricing page accessed July 28, 2026) across ticket sizes. Arithmetic only, no assumptions.

A business selling $10 items pays an effective 5.90%. A business selling $1,000 items pays 2.93%. Same processor, same contract, roughly double the rate. Average order value is a payment-cost lever, and most founders never model it as one.

---

Where does the money in a card payment actually go?

Stripe’s own documentation splits the fee three ways: interchange goes to the cardholder’s issuing bank, the assessment fee goes to the card network, and the markup goes to the payment processor (Stripe, Interchange-plus pricing explained, January 2026). Interchange is by far the largest of the three, which is why the processor’s share is a minority position.

Visa describes the same structure from the other side. Visa states that merchants do not pay interchange reimbursement fees, and that merchants instead negotiate and pay a merchant discount to their financial institution, typically calculated as a percentage per transaction (Visa USA, small business regulations and fees, accessed July 2026). Visa also frames interchange as a transfer fee between acquiring and issuing banks.

That is the four-party model in one paragraph. The merchant pays one number. The number is then split among parties the merchant never contracts with.

The only rigorously public quantification of interchange as a share of transaction value comes from the Federal Reserve’s Regulation II work on debit.

Federal Reserve measure (data year 2023)	Per transaction	As a share of value
Covered issuer, dual-message network	$0.22	0.45%
Covered issuer, single-message network	$0.24	0.52%
Exempt issuer, dual-message network	$0.62	1.41%
Exempt issuer, single-message network	$0.27	0.68%
Average covered transaction value	$48.59	n/a
Average issuer authorization, clearing and settlement cost	$0.041	n/a

Source: Federal Reserve Board, 2023 Interchange Fee Revenue, Covered Issuer Cost, and Fraud Loss Related to Debit Card Transactions (Regulation II biennial report), data year 2023. The Fed’s more recent series reports 2024 averages of $0.23 per covered transaction and $0.51 per exempt transaction (published December 19, 2025).

Two readings matter. First, an exempt-issuer debit transaction on a dual-message network consumes 1.41% of transaction value in interchange alone, which is close to half of Stripe’s entire 2.9% headline before assessments or markup. Second, the Fed measures the issuer’s own authorization, clearing and settlement cost at $0.041 per transaction against $0.22 of covered interchange. The issuing bank is the best-paid participant in the chain, and it is the one the merchant never chose.

Credit interchange runs materially higher than debit and varies by rewards tier. The networks’ schedules could not be verified line by line for this piece, so no specific credit rate is quoted; the debit figures above are the quantified anchor and credit is directionally higher.

One more public anchor is worth holding. Stripe cites The Nilson Report for the claim that US businesses paid a record $187.2 billion in card processing fees in 2024 (Stripe, Interchange-plus pricing explained, January 2026), and Nilson reports US card purchase volume of $11.903 trillion that year (Issue 1281, March 2025). Dividing the two gives an average all-in US acceptance cost of roughly 1.57% of purchase volume, against a 3.20% headline on a $100 Stripe e-commerce charge. Card-not-present retail sits at the expensive end of a market whose average is half that.

---

The Payments Take-Rate Stack: how Stripe makes money layer by layer

Every payments company can be read through the same five layers. Call it the Payments Take-Rate Stack: a waterfall from the sticker fee the merchant pays down to the spread the processor actually books, plus the software layer sold on top of the same flow.

The Stack is the analytical asset in this piece. It is deliberately reusable. Run any processor through the same five rows and the gross-versus-net question answers itself.

Layer	What it is	Who keeps it	On a $100 e-commerce charge
L1. Gross merchant discount rate	The sticker fee the merchant pays	Collected by the processor, mostly on behalf of others	$3.20 at 2.9% + 30 cents (Stripe pricing page, July 2026)
L2. Interchange	Transfer fee set by the network, paid to the card-issuing bank	The issuing bank	The largest layer. Not disclosed per transaction. Fed anchors for debit: 0.45% of value (covered) and 1.41% (exempt) on dual-message networks, data year 2023. Credit runs higher.
L3. Network assessment	Fee for use of the network rails	Visa or Mastercard	Percentage-based, set by the network. Stripe names it as a distinct component (January 2026); no verified per-transaction rate is quoted here.
L4. Net take rate	What survives as processor revenue	Stripe	About $0.40 at Sacra’s estimated 0.40% blended net take rate. Illustrative allocation of a company-level estimate to one transaction.
L4b. Attach revenue	Software priced against the same flow	Stripe	Billing 0.7% of billing volume; Connect 0.25% starting fee for platform-controlled pricing; Radar for Fraud Teams 2 cents per screened transaction; Instant Payouts 1.5%; disputes $15.00 (Stripe pricing page, July 2026)

Layer 1 and Layer 4b prices are from Stripe’s published pricing page (accessed July 28, 2026). Layer 2 anchors are Federal Reserve Regulation II debit data (data year 2023). Layer 4 uses Sacra’s estimated blended net take rate, which is a third-party estimate and not a Stripe disclosure. Layer 3 is described qualitatively because no verified rate schedule was available.

Read the Stack top to bottom and three things fall out.

Layers 2 and 3 are not costs a processor manages. They are prices set by other parties. A processor can route smarter and qualify transactions into cheaper interchange categories, but it cannot decide what an issuing bank earns.

Layer 4 is the entire payments business. Everything a payments company is valued on sits in the residual after two parties it does not control take their cut first.

Layer 4b is where the economics change. Attach products are priced against volume already flowing through Layer 1, and they carry no interchange. That is why the software layer, not the payment layer, is the margin argument.

---

Why net take rate, not the gross fee, is the real metric

Gross take rate measures how much money touches a company. Net take rate measures how much of it stays. Confusing the two is the most common error in payments analysis, and it is the same gross-versus-net gap that governs how Uber makes money on gross bookings.

Applying Sacra’s estimated 0.40% net take rate to Stripe’s published headline makes the gap concrete. On a $100 charge, the merchant pays $3.20 and roughly $0.40 survives as Stripe revenue. About $2.80, close to seven of every eight dollars collected, is pass-through.

That ratio explains behavior that otherwise looks irrational: why processors chase volume rather than price, why they accept lower headline rates to win large merchants, and why they build software businesses on the side. When you keep an eighth of what you collect, the only two paths to a larger business are more volume or more products per dollar of volume.

It also makes gross-rate comparisons close to meaningless. Two companies quoting 2.9% can run completely different net spreads depending on card mix, geography, routing quality and how much volume is negotiated. The Layer 1 to Layer 4 conversion ratio is the number that describes the business, and almost nobody publishes it.

---

Worked example: one $100 card charge, line by line

Take a single US e-commerce transaction at Stripe’s standard published pricing and follow it through the Stack. This is a worked example built from published prices and a third-party estimate; the Layer 4 allocation is illustrative because Stripe does not disclose per-transaction economics.

1. Authorization. The customer pays $100.00. Stripe’s fee at 2.9% plus 30 cents is $3.20 (Stripe pricing page, July 2026). The merchant nets $96.80.
2. Pass-through legs. Interchange leaves for the issuing bank and the assessment leaves for the network. Neither is broken out on a flat-rate account, which is precisely the point of flat-rate pricing.
3. Residual. At Sacra’s estimated 0.40% blended net take rate, roughly $0.40 of the $3.20 remains as Stripe revenue.
4. Tail risk. If that order is disputed, Stripe’s published dispute fee is $15.00, with another $15.00 for a manually countered dispute. One disputed $100 order costs the merchant the reversed sale plus a fee equal to the gross processing cost of roughly 4.7 additional $100 orders.
5. Optional acceleration. If the merchant takes an Instant Payout, published pricing is 1.5% of volume with a 50 cent minimum.

Now change one variable at a time and watch the merchant’s rate move.

Scenario on a $100 charge	Published rate	Fee	Effective rate
Domestic card, online	2.9% + $0.30	$3.20	3.20%
Manually entered card	+0.5%	$3.70	3.70%
International card	+1.5%	$4.70	4.70%
International card requiring conversion	+1.5% and +1%	$5.70	5.70%
In person via Terminal	2.7% + $0.05	$2.75	2.75%
In person, international card	+1.5%	$4.25	4.25%
ACH Direct Debit	0.8%, capped at $5.00	$0.80	0.80%

Author calculation from Stripe’s published US pricing page, accessed July 28, 2026. Arithmetic only.

The spread between the cheapest and most expensive row is more than seven times, on identical revenue. A merchant that moves a slice of volume from international cards to ACH is running a margin project, not a plumbing project. On a $1,000 invoice, ACH at 0.8% would be $8.00 but caps at $5.00, an effective 0.50%, against $29.30 on a card.

That is the practical version of the argument in why gross margin is destiny in SaaS: payment mix is a cost-of-revenue decision, and cost of revenue sets the ceiling on everything below it.

---

$1.9 trillion times a thin spread

Stripe company-announced that businesses running on it generated $1.9 trillion in total volume in 2025, up 34% from 2024, describing that as roughly 1.6% of global GDP (Stripe 2025 annual letter, published February 24, 2026). It said it serves more than 5 million businesses directly or through platforms, including 90% of the Dow Jones Industrial Average and 80% of the Nasdaq 100.

Sacra estimates Stripe’s 2025 net revenue at $6.93 billion, up 36% from $5.1 billion in 2024 and $3.82 billion in 2023 (Sacra, accessed July 2026). That estimate is not company-confirmed.

Divide the estimate by the announced volume and the implied net take rate is roughly 0.36%, or about 36 basis points. It is a coherent cross-check on Sacra’s separate 0.40% estimate, and it is the arithmetic that makes the whole model legible: a rounding-error spread applied to 1.6% of world output produces a multi-billion-dollar revenue base.

The growth rates are the more interesting comparison. Volume grew 34% while the estimated net revenue grew 36%. Net revenue growing slightly faster than volume is the signature of attach products doing work, because pure payments mix at scale normally pulls the blended rate down rather than up. One year is not a trend, and one of the two numbers is an estimate, so this is a signal to watch rather than a conclusion.

---

Is Stripe a payments company or a software company?

Both, and the software half is where the economics improve. Stripe sells Billing at 0.7% of billing volume, Connect with a 0.25% starting fee for platform-controlled pricing, and Radar for Fraud Teams at 2 cents per screened transaction (Stripe pricing page, July 2026). Each prices against volume Stripe already carries, so it adds spread without adding interchange.

That is the structural argument. Attach revenue is the highest-return revenue in a payments company because the customer acquisition already happened and the pass-through layers do not apply.

Revenue layer	Published price	Priced against	Margin character
Core card processing	2.9% + 30 cents online; 2.7% + 5 cents in person	Gross transaction value	Mostly pass-through; thin residual spread
Cross-border and FX	+1.5% international card; +1% conversion	The same transaction	Higher gross rate; carries scheme and FX cost
ACH Direct Debit	0.8%, capped at $5.00	Bank-rail transfer value	Lower price against a lower cost base than cards
Billing (Revenue suite)	0.7% of billing volume, or from $620.00 per month	Subscription volume already on Stripe	Software attach; no incremental interchange
Connect	Included at standard pricing; 0.25% starting fee for platform-controlled pricing	Platform-routed volume	Distribution layer that adds spread on existing flow
Radar	Included at standard pricing; Fraud Teams at 2 cents per screened transaction	Transaction count	Software attach priced per event
Treasury	No storage fees or minimum balance; $2.00 per wire funding	Balances and money movement	Banking-as-a-service; economics not disclosed
Issuing	10 cents per virtual card, $3.50 per physical card, $15 per dispute, plus a share of interchange	Cards created and card spend	Turns Stripe from an interchange payer into an interchange participant
Ancillary	Instant Payouts 1.5% with a 50 cent minimum; disputes $15.00	Optional events	Priced per event, at a high rate relative to the core spread

Sources: Stripe pricing page and Stripe Issuing product page, both accessed July 28, 2026. Margin character is a qualitative reading, not a disclosed figure.

The arithmetic on that table is the part worth sitting with. Consider an illustrative example using published prices and Sacra’s estimated blended net take rate: on $1 million of subscription volume already running through Stripe, the estimated net processing spread at 0.40% is about $4,000, while Billing at 0.7% adds $7,000. The software line is larger than the payments line on the identical dollar of volume, and it arrives without interchange attached.

Instant Payouts sharpen the same point. At 1.5% of volume, a single $10,000 instant payout is priced at $150, against an estimated $40 of net processing spread on that same $10,000 of volume. Optional, event-priced products can be worth several times the core spread.

The scale claims behind these lines are company marketing rather than filings, and should be read that way. Stripe says Billing recovered $8.2 billion in failed payments in 2025 across more than 350,000 companies at an average 55% recovery rate; that Connect powers more than 16,000 active platforms with over 11 million onboarded accounts; and that Radar cuts fraud 32% on average, with a 92% chance a given card has been seen before on its network. The Revenue suite was said to be on track for a $1 billion annual run rate (company-announced, February 2026), which is a product line, not company revenue.

Two products change the direction of money entirely. With Issuing, Stripe states that customers can earn a share of interchange revenue on card spend, with more than 275 million cards created (Stripe Issuing page, July 2026). Interchange stops being the cost at Layer 2 and becomes revenue at Layer 4b. With Treasury, funds sit with Fifth Third Bank, N.A. as depository partner, eligible for FDIC pass-through insurance up to $250,000 per depositor, and Stripe says businesses on the platform pay each other 4.8 million times a day (Sessions 2026, April 2026).

Pricing Billing against billing volume rather than per seat is also a deliberate choice, and it is the same structural bet examined in usage-based pricing versus seat-based pricing: revenue that grows with the customer instead of with headcount.

---

How does Stripe’s take rate compare with Adyen and PayPal?

Adyen is the cleanest public benchmark because it reports net revenue explicitly excluding interchange and scheme fees. FY2025 net revenue of 2,364.2 million euros on 1,394.3 billion euros of processed volume works out to roughly 17 basis points, at a 53% EBITDA margin (Adyen H2 2025 results, February 2026). PayPal and Block confirm the same pattern from the income-statement side.

Company (FY2025)	Volume processed	Revenue surviving pass-through	Implied net spread	Profitability marker
Adyen	1,394.3 billion euros	2,364.2 million euros net revenue	About 0.17% (author calculation)	EBITDA 1,245.7 million euros, 53% margin
Stripe (private)	$1.9 trillion, up 34% (company-announced)	$6.93B net revenue (Sacra estimate, not company-confirmed)	About 0.36% (author calculation); Sacra separately estimates 0.40%	Not disclosed
PayPal	Not used here (see methodology)	Gross profit $15,465M on revenue $33,172M	Not computed	Net income $5,233M; 46.62% gross margin
Block	Not used here	Gross profit $10,417M on revenue $24,194M	Not computed	43.06% gross margin, distorted by bitcoin revenue

Sources: Adyen N.V. H2 2025 financial results press release (February 2026); Stripe 2025 annual letter (February 24, 2026, company-announced); Sacra Stripe research page (accessed July 28, 2026); PayPal Holdings and Block, Inc. Forms 10-K for FY2025 as tabulated by StockAnalysis.com (accessed July 28, 2026).

Three conclusions survive the comparison.

Thin does not mean weak. Adyen runs roughly 17 basis points of net take and posts a 53% EBITDA margin, up from 50% in 2024. A low net take rate paired with an enterprise mix and disciplined cost structure is a very good business.

Even at the revenue line, roughly half is cost. PayPal keeps 46.62 cents of gross profit per revenue dollar and Block keeps 43.06 cents, and PayPal’s revenue is already net of some pass-through. Block’s consolidated figure is distorted by near-zero-margin bitcoin revenue, so it should not be read as a processing margin.

The Stripe row is the weakest row in the table. Two of its three numbers are company-announced or estimated. Every other row traces to an audited filing or a reported results release. That asymmetry is the honest state of the evidence, and it is why the comparables matter.

---

The mix problem: why a falling blended take rate can be good news

Blended take rate falls as a payments company succeeds. That sounds like decay and usually is not.

Four forces push the blended number down at scale. Enterprise merchants negotiate interchange-plus rather than flat pricing, and Stripe states that enterprise and high-volume businesses almost always ask for it (Stripe, Interchange-plus pricing explained, January 2026). Platforms on Connect set their own end-merchant pricing. International expansion brings local payment methods with different cost bases. And debit is structurally cheaper than rewards credit, so a shift in card mix moves the blended rate without anything else changing.

A company winning large accounts will therefore show a declining blended rate while its net revenue grows. The metric that separates the two stories is the relationship between net revenue growth and volume growth. If net revenue grows at least as fast as volume, the mix shift is being offset by attach. If it grows materially slower, the company is buying volume with price.

That is a subtler read than the take-rate headline, and it is the same discipline of reading segment mix instead of blended averages that runs through how Shopify makes money on its own take-rate engine.

---

Where this is vulnerable

Four live threats sit under the Stack, and they do not all point the same direction.

Interchange regulation and litigation cut both ways. Compressing Layer 2 reduces the merchant’s total cost, which is good for merchants and neutral-to-positive for a processor’s competitive position. It also removes cover: when the pass-through shrinks, the processor’s markup becomes more visible and more negotiable. A world with cheaper interchange is a world where Layer 4 gets scrutinized harder.

The networks own the rails. Visa and Mastercard set interchange schedules and assessments. A processor optimizes routing and qualification within rules it does not write. Any structural change in network economics arrives as an exogenous shock to Layer 4.

Stablecoins disintermediate the rail Stripe monetizes. Stripe reported that stablecoin payments volume roughly doubled to around $400 billion, with about 60% estimated to be business-to-business, and that Bridge volume more than quadrupled (Stripe 2025 annual letter, February 2026). Stripe is building the alternative rail itself, through Bridge and the Tempo chain announced at Sessions 2026. That is a hedge and a self-inflicted wound at once: a payment that settles on a stablecoin rail does not generate interchange, which is good for merchants and structurally different for a company whose spread sits on card economics.

The checkout surface is moving. Stripe announced 288 launches at Sessions 2026, including an Agentic Commerce Suite extended with Google across AI Mode and the Gemini app, Link wallets for agents using one-time-use cards, and streaming payments combining usage tracking with stablecoin micropayments (Stripe Sessions 2026, April 29, 2026). If purchases increasingly originate inside an AI assistant, whoever controls that surface gains pricing power over whoever processes the payment. Owning the checkout integration early is a land grab against exactly that risk.

There is a fifth vulnerability that is analytical rather than commercial. Every distribution-controlling intermediary eventually faces pressure on the rate it charges, which is the story of Apple’s App Store economics under pressure. Payments has a structural defense App Store fees lack, because most of the fee genuinely leaves the building. That defense is only as strong as the industry’s ability to explain it.

---

The bear case: what the skeptics get right

The strongest argument against this framing is not that any figure is wrong. It is that a roughly 40 basis point spread is a thin moat, and the evidence for it is thinner than the argument requires.

A competitor already runs at less than half the spread, profitably. Adyen converts 1,394.3 billion euros of volume into 2,364.2 million euros of net revenue at a 53% EBITDA margin (Adyen H2 2025 results, February 2026). If an enterprise-focused competitor is structurally content at roughly 17 basis points, then a 36 to 40 basis point blended rate is not a defended position, it is a mix outcome that competition can compress. The bear reads Stripe’s higher estimated spread as evidence of a small-business-heavy book, not of pricing power.

The software attach may not be large enough yet. Stripe said the Revenue suite was on track for a $1 billion annual run rate (company-announced, February 2026). Against a Sacra-estimated $6.93 billion of net revenue, that is a meaningful but not dominant line, and it is the one attach figure Stripe quantified at all. Connect, Radar, Treasury, Issuing and Capital revenue are described in customer counts and marketing claims, not dollars. The attach thesis rests on products whose revenue nobody outside the company can size.

The valuation gap is unexplained. The most recent tender valued Stripe at $159 billion (company-announced, February 2026), while Sacra reports a January 2026 internal 409A mark of $106.7 billion. Both can be defensible under different methodologies, and neither is auditable from outside. This is an observation about disclosure, not a view on any security.

There is no filing. No audited income statement, no segment disclosure, no cohort economics, no churn, no gross margin. Everything favorable in this article traces to a company blog post or a research firm’s estimate. An analyst applying the standard in how to read a tech S-1 like an operator would find almost none of the required inputs present.

Weighing it honestly: the bear case does not break the model, it bounds the confidence. The Stack is structurally correct because Stripe, Visa and the Federal Reserve all describe the same waterfall. What the bear case correctly denies is precision. Anyone quoting Stripe’s net take rate to two decimal places is quoting an estimate, and anyone treating the software attach as proven is treating marketing as accounting.

---

What operators should take from this

The transferable lesson is not about Stripe. It is that your payment cost is a curve you control more of than you think, and almost nobody measures it correctly.

Here are six concrete moves.

1. Compute your true effective rate, monthly. Divide total payment costs by total processed volume, including disputes, refunds, currency conversion, instant payouts and platform fees. That number, not the sticker rate, is what belongs in cost of revenue. Most operators quote 2.9% and actually pay something higher.
2. Fix average order value before you negotiate the rate. The table above shows 5.90% at a $10 ticket against 2.93% at $1,000, purely from the fixed 30 cents. Bundling, annual plans and minimum order sizes move payment cost more than any negotiation will at small scale. In the consumer subscription products I operate, the fixed per-transaction component is the most underrated line in the pricing model, because it turns a percentage fee into a regressive tax on the cheapest tier.
3. Know when interchange-plus is worth asking for. Stripe states that enterprise and high-volume businesses almost always request it (January 2026). Interchange-plus makes Layers 2 and 3 visible on your statement, so you can verify the pass-through instead of trusting a blended number. Ask when you have the volume to be worth re-papering and the finance capacity to read a line-item statement.
4. Price disputes and fraud into gross margin, not into a footnote. At Stripe’s published $15.00 dispute fee, one disputed $100 order costs the gross processing fee of roughly 4.7 additional $100 orders, plus the reversed sale. Model a dispute rate per cohort and put it in cost of revenue where it belongs.
5. Run every attach product as a rate increase and test it as one. Billing at 0.7% of billing volume is a real cost against a real benefit, and Stripe’s own claim is that businesses recover 55% of failed payments on average. Compute the recovered revenue against the 0.7% and the build-versus-buy alternative, then decide. Do the same for Radar tiers, Instant Payouts and Connect fees.
6. Read the contract for everything that is not the rate. Instant payout pricing, dispute counter fees, currency conversion, reserve and rolling-hold terms, chargeback thresholds, and the cap structure on bank-rail products. On a $1,000 ACH transfer the published 0.8% would be $8.00 but caps at $5.00, and cap structures like that are where real money hides.

The meta-move is the Stack itself. Before you compare two processors, write down what each one collects at Layer 1, what you can see of Layers 2 and 3, and what you are paying at Layer 4b. A quote that looks cheaper at Layer 1 and expensive at Layer 4b is a worse deal that reads as a better one.

---

Methodology: how the computed figures were derived

• Inputs. Stripe’s pricing page and product pages (accessed July 28, 2026); Stripe’s 2025 annual letter and Sessions 2026 announcement (company-announced); Sacra’s Stripe research page (third-party estimates); Federal Reserve Regulation II data (2023 and 2024); The Nilson Report Issue 1281 (March 2025); Adyen’s H2 2025 results; PayPal and Block Forms 10-K for FY2025 as tabulated by StockAnalysis.com.
• Calculations. Effective-rate tables are arithmetic on published prices. Stripe’s implied net take rate is $6.93B divided by $1.9T, about 0.36%, combining a third-party revenue estimate with a company-announced volume figure. Adyen’s is 2,364.2 million euros divided by 1,394,300 million euros, about 0.17%. US average acceptance cost is $187.2B divided by $11.903T, about 1.57%.
• Assumptions. That a blended net take rate can be applied illustratively to one standard-priced transaction. Real per-transaction economics vary with card type, geography, routing and contract.
• Sensitivity. If Stripe’s true blended net take rate is 0.30% rather than 0.40%, the residual on a $100 charge falls from about $0.40 to about $0.30 and the attach argument gets stronger, not weaker.
• What this misses. PayPal’s total payment volume and quarterly transaction take rate could not be confirmed against a primary source, so no PayPal spread-on-volume figure is computed. No credit interchange rate is quoted. Stripe’s gross revenue, profitability, churn and segment mix are all undisclosed.

---

How the pieces fit together

Stripe’s model reduces to five moves, in order:

1. Quote a simple, high headline rate that a small business will accept without negotiation.
2. Pass the large majority of that fee through to issuing banks and networks, keeping a spread estimated in the low tens of basis points.
3. Apply that spread to extraordinary volume, company-announced at $1.9 trillion in 2025, or roughly 1.6% of global GDP.
4. Sell software priced against the same flow, Billing, Connect, Radar, Treasury and Issuing, where no interchange applies and the customer is already acquired.
5. Build the next rail before the current one is disintermediated, which is what the stablecoin and agentic-commerce work is.

The headline number tells you what Stripe collects. The Stack tells you what it earns. Those are different companies, and only one of them is the business.

---

Analysis, not investment advice. Stripe is a private company that files no Form 10-K, so all Stripe figures here are company-announced or named third-party estimates and are labeled as such inline. Public-company figures are drawn from the filings and results releases cited by company and fiscal period. Frameworks here, including the Payments Take-Rate Stack, are for understanding business models and tradeoffs, not for making buy or sell decisions.

Want the full toolkit for reading filings like this, the take-rate waterfall worksheet, the effective-rate calculator, and the attach-revenue model used above? It’s in the Tech Business Analysis Playbook.