The Same Processing Function Earns 16.96 Basis Points and 55.66. Nothing About the Software Differs.

Three companies process a payment. Each authorises it, routes it, settles it and reports on it back to the merchant. The software does the same thing, at the same quality, for the same kind of customer. And one of them keeps 16.96 basis points for the work, another 30.74, and the third 55.66.

That is a 3.28x spread on an identical function. It is not a technology gap, and it is not an efficiency gap. It is a market-structure gap, and it is the single most useful observation in the payment gateway market — because it explains why the blended take rate for the whole layer is falling while individual providers hold or raise their rate on the accounts they care about.

Four revenue lines on one transaction

The gateway layer is usually described as software. Commercially it is four revenue lines stacked on the same transaction, and the mix between them sets the take rate.

Processing fee — a rate in basis points on value processed, sometimes with a per-transaction component. Visible, and the easiest line to negotiate away. Acquirer spread — where the provider also carries the acquiring position, it keeps the difference between the merchant discount it charges and the interchange plus scheme fees it pays. Platform and subscription — fixed monthly or per-seat charges, modules, terminal rental, fraud tooling; recurring and price-insensitive. FX and payout margin — the spread on conversion and cross-border settlement, invisible to most merchants and frequently larger than the processing fee on cross-border volume.

GMO Payment Gateway publishes this split, which makes it the cleanest public decomposition of a gateway’s P&L anywhere. In its most recent reported quarter the four lines were: initial revenue JPY 1,754.2 million, stock and recurring JPY 1,723.1 million, transaction fee JPY 2,923.0 million, and merchant-acquiring spread JPY 4,731.9 million — JPY 11,132.2 million in total on consolidated processed value of JPY 2.0 trillion and 0.97 billion transactions.

Read that ratio carefully. The transaction-fee line alone works out to roughly 14.6 basis points. The blend, with the acquiring spread included, is 55.66. The acquiring spread is approximately three times the size of the processing fee. Two companies can be competing for the same merchant, doing the same technical work, and one of them earns four times as much — because it is standing in a different place in the chain.

Three regimes, three bands

Interchange is the fee an acquirer pays an issuer on a card transaction, and it is the largest single component of a merchant discount. It is also the boundary of this market: it accrues to the issuing side, so every change to it moves the merchant’s cost without moving the gateway’s revenue — except where the gateway is also the acquirer, in which case the spread it earns is a residual over interchange and moves with it.

The world divides into three interchange regimes, and they produce the three bands.

Europe capped interchange by law, which compresses the merchant discount and with it the absolute spread available inside a domestic card transaction. The processing fee there has to be earned on service rather than on spread. Japan disclosed it — METI, working with the Japan Fair Trade Commission, published the standard rates for credit card interchange in 2022 and the allocation of merchant fees in 2023. Nothing was capped. What changed is that a merchant can now see which party to negotiate with, and a domestic acquirer earning a residual over a now-published number has a visible ceiling. The United States left it to scheme rules and partial debit regulation, which is why the widest spread sits there.

The mechanism is mechanical: a gateway’s take rate is what remains of the merchant discount after interchange and scheme fees. Where interchange is capped, the remainder is small and must be defended by service. Where it is disclosed, the remainder is visible and must be defended in negotiation. Where it is unpriced, the remainder is wide and attracts entrants.

The rail underneath is the actual threat

The rule that matters most to this market is not an interchange rule. It is the European Instant Payments Regulation, because it attacks the card transaction from the rail beneath it.

Euro-area payment service providers had to be able to receive instant credit transfers from 9 January 2025, and to send them with verification of payee from 9 October 2025. Electronic money institutions and payment institutions follow on 9 April 2027, and non-euro-area member states on 9 July 2027. Charges for instant transfers may not exceed charges for standard credit transfers.

Three design choices in that regime do three different things. Mandatory reachability makes the rail universal, so a merchant cannot be told it is unavailable. Price parity makes it permanently cheaper than a card by statute rather than by competition. And verification of payee — the name-check before a transfer is sent, returning match, close match or no match — removes the main operational objection to using it. When a rail is mandated, priced at parity and equipped with a name check, the case for routing a domestic payment through a card collapses from both ends at once.

Only two stages are contestable by software

Issuance, clearing and acquiring are licence- and balance-sheet businesses. That single fact explains an enormous amount of otherwise puzzling corporate behaviour: it is why every serious gateway eventually tries to own acquiring, because acquiring is where the spread sits, and it is why the pure processing fee compresses toward its own cost.

What remains above it — orchestration, fraud, reconciliation, payouts, terminals — carries the highest gross margin in the chain and the most defensible position, because it owns a measured outcome rather than a price. That is the arbitration the whole sector is running: out of the commoditising line and into the one where the merchant cannot compute an alternative.

But price is not what merchants choose on

This is where the take-rate series misleads. Merchants do not pick a provider on basis points, and the primary research behind this analysis is consistent about it across six markets: cost is negotiated annually, authorisation is reviewed monthly, and the monthly review is what gets a provider replaced.

Three things decide an account, in the order merchants describe them. Authorisation uplift — the share of attempted payments that complete, measured by method, issuer country and device. Coverage of the methods that actually clear — in a market where 82.7 per cent of cashless value is card and 10.2 per cent is code, the provider that settles both on one contract beats the provider that is cheapest on card. Operational depth — settlement timing, chargeback handling, reconciliation, the quality of the integration team. Unglamorous, and the most common reason an incumbent loses an account.

The merchant’s real cost is not the fee. It is the fee multiplied by the share of attempted payments that are declined, plus the operational cost of recovering them. A provider that raises authorisation by one percentage point on a three per cent margin business is worth more than one that cuts its rate by five basis points.

Which is why a market can show a 25.5 per cent decline in the blended take rate over eight years without the competition that produces it ever being fought over price. The rate falls because the mix changes. The individual provider still holds its rate on the accounts it wants.

For the size and shape of that change across the layer as a whole — revenue reaching USD 60,307.3 million by 2033 at 6.94 per cent — see our Global Digital Payment Gateway Market report, and for the country where the 3.28x spread is widest, our Japan Digital Payment Acceptance coverage.

Method note: figures from Global Digital Payment Gateway Market Report 2026–2033. Take rates are calculated from each provider’s own reported net revenue against its own reported processed volume — Adyen, Stripe and GMO Payment Gateway are the three public disclosures that make this layer observable. Primary research establishes how the buying decision is made; it carries no number. No third-party market-research estimates are used or cited.