Inside the ATM

The fascia is the machine as far as almost everyone is concerned: a cabinet, a screen, a slot, a brand. It is also the thinnest claim in the whole assembly. The research behind this page went looking for who builds each module inside an ATM and kept arriving at the same answer — a handful of specialist suppliers, most of them Japanese, most of them invisible, none of them the company whose name is on the front.

What the cabinet genuinely is: a weather envelope, a security boundary, and an ergonomic problem. It has to survive a car park, resist somebody attacking it with hand tools long enough for a response, and still let a person of any height reach the keypad. The five generations of shapes in this gallery are five different answers to that, from Docutel's 1973 upright box to a modern frame built around a portrait touchscreen.

The first cash machines had nothing worth calling a display. What replaced it — a CRT flanked by function display keys — became so completely the idea of an ATM that the layout survived thirty years and outlived the tube itself.

Two constraints shaped everything after that, and neither is obvious from a desk. An ATM screen is read outdoors, so sunlight readability is a purchasing specification rather than a nicety, and a whole aftermarket exists to retrofit brighter panels into installed machines. And when touch replaced physical keys, it broke the machine for blind users, which is why a patent for a tactile touch screen — one you can feel your way around — sits at the top of this gallery.

The card slot is the most concentrated part of the whole supply chain. Sankyo Seiki — Nidec Sankyo from 2005, Nidec Instruments from 2023 — built its first magnetic card reader in 1971 and aimed it at cash machines. It now claims about 80% of the readers built into the world's ATMs. Part-number archaeology across three competing bank-ATM brands keeps surfacing the same Sankyo motorised family behind all of them.

It is also where the security arms race is most visible in the mechanism itself. A skimmer is a reader taped to the throat of the real one; the counter-measure is to move the card unevenly as it goes in, so the parasite gets an unreadable signal. That is a jitter drive, and there is a patent for it. Later readers add shutters against fishing, encrypted read heads, and contactless.

The keypad is the part that had to be invented before any of the rest made sense. Somebody had to decide that a short number a person could remember was enough to authorise a stranger's machine to hand over cash. The patent for that idea is in this gallery, filed in 1967 by Anthony Davies and James Goodfellow at Chubb.

Everything since has been about making the pad itself untrustworthy to attack. A modern EPP is a tamper-responsive module: drill it, freeze it, or lift its lid and it destroys its keys rather than reveal them. It is certified under PCI PTS, and the certification expires, which quietly forces hardware refreshes across entire estates on a schedule set by a standards body rather than by the bank.

Receipt printing is the exception that makes the rest of the supply chain legible. Where card readers and note handling went out to Japanese specialists, the integrators largely kept receipt printers: NCR shared engineering with its retail point-of-sale printer line, Wincor Nixdorf ran its own needle and thermal families, Diebold engineered its own.

The technology arc is short and total — impact, then dot matrix, then thermal, with two-sided thermal arriving late to halve the paper an estate consumes. The interesting survival is impact printing, which held on far longer than it should have in exactly one place: passbooks, which have to be struck rather than heated.

This is the hard mechanical problem, and the one the industry has never fully brought in-house. De La Rue was feeding and counting notes mechanically before anyone put a machine outside a bank, and it supplied the dispensers other people's ATMs were built around — a lineage that ran through Talaris and ended inside Glory. Notes are limp, dirty, stuck together and inconsistent between countries; a picker that works on crisp US dollars is not a picker that works on old rupees.

The generational break is recycling. A recycler takes deposits and pays the same notes back out, which collapses the deposit module and the dispenser into one machine and cuts the cash-in-transit runs that dominate the running cost of an estate. Japan got there first — OKI in 1982 — and Western makers mostly bought the module in rather than build it. Wincor Nixdorf is the notable exception that kept recycling in-house.

For the first two decades the controller ran a proprietary real-time operating system on custom hardware, which is an underrated security property: there was nothing to attack with, because there was no general-purpose computer in the machine. That ended with the move through OS/2 to Windows, and the arrival of CEN/XFS — a standard interface to the dispenser, reader and pad that finally let a bank run one application across several vendors' hardware.

XFS is the hinge. It made multi-vendor estates possible, and it made the black-box attack possible, because it defined a documented way to ask the dispenser for money. The countermeasure is for the dispenser to stop trusting the PC above it and cryptographically authenticate every instruction — which is what the anti-black-box patents describe.

The journal is the ATM writing down what happened. When a customer said the machine had not paid out and the bank said it had, the paper roll inside the machine was what settled it, which made the journal printer a legal instrument as much as a peripheral — and made a jammed one a genuine problem.

It went electronic, and the physical module largely disappeared into storage and the network. What did not disappear is the underlying question of whose record is authoritative, which is the same question the anti-black-box work is answering in a different register.

The oldest technology in the machine, and the one with the longest corporate memory. The manufacturing registry behind the companion globe reaches back to Chubb's Wolverhampton works in 1818 and Sargent & Greenleaf from 1857, and those names are in an ATM story because the bottom of a cash machine is a rated safe with a certified lock on it — a specialism with its own makers, its own standards and its own century of practice.

Two things drive its evolution. Locks became electronic and auditable, with one-time combinations issued centrally so a cash-in-transit crew can open a specific safe once, at a specific time, without anyone holding a reusable secret. And attackers escalated: Europe's gas-attack epidemic produced safes designed with blow-out vents, engineered on the assumption that the safe will be filled with explosive gas and that the job is to survive the blast rather than prevent it.