🏧 ATMs

The economics of a withdrawal: who pays what, and in which direction

The economics of an ATM withdrawal come down to how costs and revenue are split among the three parties to the transaction: the cardholder, the institution that issued the card, and the ATM operator. The operator is not always a bank. The cash comes out of a cassette owned by the operator, but the account debited is held by the issuer. Money therefore flows in the opposite direction from the cash, to the operator that dispensed it. Depending on whether the issuer or the cardholder pays, the cost of the withdrawal falls on the customer’s bank or on the customer.

Withdrawal interchange is the fee the issuer pays the operator on each transaction. It pays for a service provided to the issuer’s customer by a machine the issuer does not own. In a card payment, interchange flows the other way, from the merchant’s side to the issuer. The opposite directions reflect what the customer receives. A card payment delivers an authorization, whereas a withdrawal delivers banknotes that had to be bought, insured, and transported in advance. Withdrawal interchange covers those physical costs, which card payments do not incur.

The UK’s LINK network publishes its pricing in detail, which makes the UK one of the few markets in Europe with documented rates. LINK’s board sets the rates each year based on a cost study of ATM operators. For 2026, standard interchange is 26.5p per withdrawal at an off-site ATM and 24.4p at a branch ATM. Non-cash transactions, such as balance inquiries and PIN changes, pay 16.8p and 12.6p respectively (LINK, Interchange Report 2025).

The same study puts the average cost of a transaction at an off-site ATM at 39.8p. Hardware is a small part of it. Purchase and installation account for 2.0p, site rental and local taxes for 4.1p, and maintenance for 4.2p. The cash itself costs 21.6p: 11.4p is the opportunity cost of idle cash, and 10.2p covers purchasing, delivery, handling, insurance, and theft (LINK, Cost Study 2025).

The second funding model puts the cost on the cardholder, who pays it directly at the ATM. This surcharge goes to the operator. It is disclosed on screen, and the cardholder must accept it before the cash is dispensed. The US market uses it across the board, and the cardholder’s bank usually adds an out-of-network fee. The cardholder then pays twice for one transaction. In 2025, the average surcharge was $3.22 and the average out-of-network fee $1.64, so a single withdrawal cost $4.86 (Bankrate, Checking Account and ATM Fee Study, 2025).

International withdrawals can carry a third charge. With dynamic currency conversion, the cardholder is debited in their home currency at the operator’s exchange rate rather than the network’s. The markup goes to the operator and its conversion partner, with no cap. Regulation (EU) 2019/518 amended Regulation (EC) No 924/2009 on exactly this point. Since April 19, 2020, the markup must be displayed as a percentage over the European Central Bank’s reference rates before the cardholder chooses. Yet the ATM screen preselects one of the two currencies, and that default steers the choice.

Who deploys ATMs: bank fleets, independent deployers, shared fleets

Deployment means installing and operating ATMs, which covers owning the hardware, loading cash, and paying for the site. Three types of players do it, each with its own economics. Retail banks treat their ATM fleet as a customer service and an extension of the branch. For independent ATM deployers, the ATM is the product, since interchange and surcharges are their revenue. Shared-fleet companies are set up by competing banks to take their ATMs off their balance sheets without handing them to a third party.

An independent deployer’s income statement has few line items. It earns a per-transaction fee set by a network or an industry-wide agreement, and it pays site rent, the cost of cash, and the cost of cash-in-transit runs. Its margin therefore depends on the number of withdrawals per machine, a variable it does not control. When transaction counts fall, costs do not fall in proportion, since they are mostly fixed per machine and per run.

Euronet Worldwide publishes its financial statements, which gives a sense of scale. The group reported 56,818 installed ATMs as of December 31, 2025, up 3% year over year, of which 50,959 were active. Its electronic processing segment generated $1,283.7 million in revenue for the year and $278.8 million in operating income (Euronet Worldwide, full-year 2025 results, February 12, 2026). That is an operating margin of 21.7%. These figures describe a volume business, whose earnings come from the number of transactions processed rather than the margin on each one.

The Belgian arrangement is documented in detail because a competition authority examined it piece by piece. Batopin brings together BNP Paribas Fortis, KBC, ING, and Belfius, which replaced their four ATM fleets with a single network under the neutral CASH brand. The initial plan called for 750 sites and 2,240 ATMs, less than half as many machines as the four networks combined (Belgian Competition Authority, Batopin Q&A, March 2025). Cutting the number of machines was the whole point of the joint venture.

The Belgian Competition Authority opened an investigation into the joint venture, then closed it on March 25, 2025, by making commitments binding. Its analysis focused on how far residents would have to travel. It divided the country into about 33,000 one-square-kilometer cells and placed a representative user at the center of each. It then compared the distance to the nearest site in two scenarios, with and without the joint venture. Its preliminary finding was a significant decline in accessibility and a risk of heavier usage per machine.

The binding commitments set numerical targets that Batopin must meet through the end of 2030. The network will add 70 sites by the end of 2027, on top of those already planned under the March 2023 agreement with the federal government. It must put an ATM within 5 kilometers of home for 95% of Belgians, and a deposit-taking machine within 5 kilometers for 85%. Withdrawal availability must not fall below 95%. Batopin reports on these targets to the Belgian Competition Authority each year.

Coverage and its decline: where ATMs are disappearing, and at what cost

Coverage refers to how ATMs are distributed across an area and how far people are from the nearest one. Its decline is measured by the number of machines in service and by the change in that distance. The euro area had about 248,900 ATMs at the end of 2025, down 1.3% year over year. It had about 249,300 in mid-2025, when the decline was measured at 2.9% (European Central Bank, payment statistics for the second half of 2025, published July 22, 2026). Contactless acceptance is spreading faster than the fleet is shrinking: 38% of ATMs accepted it at the end of 2025, up from 34% six months earlier. The decline is slow and steady, and it is unevenly spread across locations.

UK figures show how that decline is distributed. The free-to-use fleet fell from 37,299 to 35,468 ATMs in 2024 alone, a drop of 4.9%. Branch ATMs fell 6.0% and off-site ATMs 4.4%, the difference reflecting bank branch closures (LINK, Interchange Report 2025). At the end of 2025, the country had 33,710 free ATMs and 8,693 pay-to-use ones, down 5% year over year (LINK, 2026). A branch ATM disappears with the branch that houses it, which explains the gap between the two rates.

France is on the same trend, starting from a denser network. Its fleet stood at 40,804 ATMs at the end of 2025, after about 1,700 machines were removed during the year, a 4.0% decline. The cuts were concentrated in dense urban areas, where several ATMs competed on the same sidewalk. Population coverage did not decline as a result. At the end of 2025, 98.6% of the population was still within a 15-minute drive of a site with an ATM. Including cash access points at merchants, that share reached 99.8% (Comité national des moyens de paiement, France’s national payments committee, July 2026).

The cost of a withdrawal depends on where the ATM is. A rural ATM needs the same cash-in-transit run as an urban one for a fraction of the withdrawals, which raises its cost per transaction. LINK’s answer was a tiered rate rather than an outside subsidy. So-called protected ATMs, located 1 kilometer or more from the next free ATM, receive higher interchange: 30.5p versus 26.5p for an off-site withdrawal in 2026. About 3,600 of 24,700 machines also receive a volume premium of up to £2.75 per withdrawal. Issuers therefore fund the extra cost of isolated locations through network interchange.

The program accounts for a small share of the money flowing through the network. Protected interchange and premiums totaled £17.7 million in 2025, or 5.9% of the network’s total interchange. Protected ATMs numbered 3,610 in April 2025 and 3,734 in April 2026, 11.4% of the country’s free ATMs (LINK, Interchange Report 2026). LINK’s board expects about 18,000 free ATMs in 2032 and puts the minimum consistent with its commitments at about 10,000. The 2032 forecast is still above that floor.

Alternatives to the ATM: merchant checkouts, cardless withdrawals, service points

An ATM alternative is any arrangement that provides access to cash where the ATM fleet is shrinking. Three exist today. Customers can withdraw cash at a merchant’s checkout, withdraw without a card at an existing ATM, or use an in-store service point that also takes deposits. The first shifts the burden to the retailer, the second only changes how the customer authenticates, and the third turns the store into a mini branch. A count that lumps them together overstates the coverage they provide.

Cash back at the merchant means the payee of a payment hands cash to the payer. Article 3(e) of Directive (EU) 2015/2366 excludes this from the scope of payment services. The exemption applies only if the cash comes with a purchase of goods or services and the customer asked for it just before the transaction. A merchant that hands out cash as part of a sale is therefore not carrying out a regulated activity. One that does so without a sale falls outside the exemption. What defines the merchant’s regime is whether there is a purchase, not the act of handing over cash.

The proposed Payment Services Regulation lets merchants provide cash without a purchase, without a payment institution license, subject to a cap and with customer authentication. The European Commission set the cap at €50 in its 2023 proposal. A provisional political agreement was reached on November 27, 2025, but the text had not been published in the Official Journal at the time of writing. The binding cap will be the one in the published text, and any figures cited before then are working numbers.

This channel already works in the countries that measure it. France had 30,051 private cash access points at merchants at the end of 2025, up 5.5% year over year. These points lift coverage from 98.6% to 99.8% of the population (Comité national des moyens de paiement, July 2026). In the euro area, 57% of consumers say they can withdraw cash at a store checkout (ECB, SPACE 2024). That figure measures an option consumers report, not withdrawals actually made at checkouts, which remain far below that potential.

Yet ATMs still handle most of the volume. The LINK network recorded 1.272 billion withdrawals in 2025, across a fleet of 42,403 ATMs, 33,710 of them free (LINK, Statistics and Trends, 2025). The free fleet shrank 5% over the same period, and withdrawals are shifting to the remaining machines. A cash-in-transit run serves a site regardless of how many withdrawals it handles. A smaller fleet therefore reduces cash logistics costs less than proportionally.

A cardless withdrawal is one in which the cardholder authenticates with something other than the physical card. There are two approaches, and only one is interoperable at scale. In the first, the issuing bank’s app handles authentication and provides a code to enter at one of the bank’s own ATMs. The second relies on a national scheme that makes cardless withdrawals work at every connected ATM. India built this second approach with NPCI’s UPI-ATM, which displays a dynamic, single-use QR code on the ATM screen. The customer scans it with their UPI app and confirms with their PIN, up to ₹10,000 per transaction (NPCI).

Contactless offers a third approach. A cardholder who taps a phone on the ATM authenticates on the device, without inserting a card or exposing the reader to a skimming device. Fleets are being upgraded slowly: 38% of euro area ATMs accepted contactless at the end of 2025, up from 34% six months earlier (ECB). A cardless withdrawal program that requires a fully equipped fleet therefore reaches no more than two ATMs in five. Equipment rates vary by operator, so the European average says nothing about any given fleet.

• Check the legal regime before signing. Cash back with a purchase falls under the Article 3(e) exemption in PSD2. Cash without a purchase falls under the new regime in the EU payments package, with the national cap to be confirmed once the text is published.
• Put in writing who keeps the till stocked. A merchant that hands out cash depends on its cash takings. At a mostly card-based point of sale, running out is structural, not accidental.
• Pay merchants for the service, or accept that they will drop it. Checkout time is a real cost for the merchant, and an unpaid program dies as soon as lines get longer.
• Track unavailability. Without reporting when a limit is reached or the till is empty, the coverage shown on a service map is meaningless.
• Distinguish withdrawals from deposits. A cash withdrawal point does not replace a deposit ATM for a merchant that needs to bank its takings, and the Belgian commitments set separate thresholds for the two, 95% and 85%.
• Measure access in travel time. Authorities think in road distance or minutes, and a count of access points does not answer their question.

Cash access obligations: who must comply, with what, and under whose oversight

A cash access obligation is a rule that requires designated players to keep cash withdrawals available across a given area. These obligations emerged once withdrawals stopped being profitable everywhere. They take three forms, which differ in who is bound and which authority enforces them. Legislation designates institutions and requires them to fill identified gaps. An industry-wide agreement commits a whole sector to numerical targets, overseen by an authority. A competition decision makes an operator’s commitments enforceable, with a similar effect and no new legislation.

The UK chose legislation. The Financial Services and Markets Act 2023 gave HM Treasury the power to designate institutions and the Financial Conduct Authority the power to supervise them. The rules are set out in policy statement PS24/8, published on July 24, 2024, and have applied since September 18, 2024. The government has designated 14 banks and building societies. Institutions that are not designated fall outside the regime.

The UK mechanism relies on local assessments rather than a national quota. A designated institution must assess local access before any closure. It must respond to assessment requests from residents or local organizations, then provide a reasonable additional service if a significant gap is found. It cannot close an existing service until the replacement is open. The replacement can be a shared banking hub, an ATM, including one that takes deposits, or a service at a post office.

Belgium has layered two instruments whose effects add up. The March 31, 2023, agreement between the federal government and Febelfin binds issuers and operators until December 31, 2027. It requires 207 additional sites, then 80 more ATMs in the Brussels-Capital Region and in urban areas with more than 2,500 residents per ATM. Every municipality must keep at least one ATM. The National Bank of Belgium measures the 95% availability target monthly. The commitments made binding by the competition authority in 2025 then extend coverage through the end of 2030.

EU law addresses the issue in two separate texts aimed at different parties. The payment services package targets operators. It requires registration for ATM deployers that do not hold payment accounts, without requiring full authorization, and it imposes fee transparency. The regulation on the legal tender status of euro banknotes and coins targets member states. It requires each state to ensure sufficient access to cash across its territory and to report to the European Commission and the European Central Bank. Both texts are still under negotiation.

Oversight of these regimes relies on data reporting, whose frequency and content vary from one authority to another. The National Bank of Belgium measures availability monthly, while the Belgian Competition Authority receives an annual report on site rollout. The Financial Conduct Authority handles local requests, and the Banque de France publishes an annual report on ATM coverage. An operator entering one of these markets therefore faces reporting obligations before coverage obligations. What an authority can demand depends on the data it receives.