The first ATM was installed outside a Barclays branch in Enfield, north London, on 27 June 1967. The customer inserted a paper voucher, entered a PIN, and received ten pounds. That machine represented the state of the art in banking technology for nearly two decades. Today, a customer in Singapore can open a bank account, fund it, and begin investing in government bonds through a mobile app in under three minutes, using facial recognition for identity verification and real-time API connections to securities markets. The distance between those two moments is not just technological. It is architectural. The global banking-as-a-service market that powers today’s banking technology reached $18.6 billion in 2024, according to Global Market Insights, growing at 15.1% annually.
The Mainframe Era: 1960s to 1990s
Banks were among the first commercial adopters of computing. Bank of America deployed ERMA (Electronic Recording Machine, Accounting) in 1959, the first system to automate cheque processing using magnetic ink character recognition (MICR). By the mid-1960s, every major bank in the US and Europe was installing IBM mainframes to process transactions, maintain ledgers, and generate regulatory reports.
These systems were built to last. COBOL, the programming language designed for business data processing in 1959, became the standard for banking software. The mainframes that banks installed in the 1970s and 1980s still process trillions of dollars in daily transactions at many of the world’s largest institutions. Reuters estimated in 2022 that 95% of ATM transactions still pass through COBOL code.
The Boston Consulting Group projects fintech revenues will reach $1.5 trillion by 2030, with embedded finance and digital lending accounting for the largest share of projected growth.
According to CB Insights’ 2024 fintech report, global fintech funding declined 40 percent between 2022 and 2024, pushing the sector toward consolidation and a sharper focus on profitability over growth at all costs.
The defining characteristic of mainframe-era banking technology was batch processing. Transactions entered during the day were accumulated and processed overnight. Account balances updated once per day. A deposit made at 2pm on Monday would not appear in the customer’s balance until Tuesday morning. This was not a limitation anyone noticed until real-time alternatives emerged.
Internet Banking and the Channel Layer: 1995 to 2015
The internet introduced a new channel for banking but did not change the underlying technology. Banks built web portals that displayed account balances and allowed basic transactions (transfers, bill payments). These portals connected to the same mainframes that powered the branch network. The data was the same. The processing was the same. Only the access point was different.
Mobile banking followed the same pattern. When banks launched mobile apps starting around 2008, the apps were thin clients that communicated with the same backend systems. A customer checking their balance on a mobile app in 2012 was, in most cases, seeing data processed by a mainframe built in 1985.
This approach had a practical advantage: it minimised risk. Banks did not need to rebuild their core systems to offer internet or mobile access. They added a presentation layer on top. But it also meant that the fundamental constraints of batch processing, limited API capability, and slow product development persisted underneath the modern-looking interface.
The Cloud-Native Shift: 2015 to Present
The current era of banking technology is different in kind from what came before. Rather than adding new interfaces to old systems, a new generation of companies is building banking infrastructure from the ground up on cloud platforms.
Thought Machine (founded 2014) built Vault, a core banking platform that runs on Google Cloud and processes transactions in real time. Mambu (founded 2011) built a composable banking platform on AWS. 10x Banking (founded 2016 by former Barclays CEO Antony Jenkins) built a cloud-native core specifically to replace the mainframes he had struggled to modernise during his time running a traditional bank.
These platforms share four characteristics that mainframes lack. They process transactions in real time, updating balances within milliseconds rather than overnight. They expose every function through APIs, enabling integration with third-party services in days rather than months. They scale automatically with demand, eliminating the capacity planning that mainframe operations require. And they use modern programming languages that today’s engineers know how to work with.
Cloud deployment now accounts for 67% of the banking-as-a-service market, per Global Market Insights. Platform-based models hold a 69% share. These numbers reflect a technology generation change: the industry is moving from proprietary, on-premise mainframes to shared, cloud-based platforms that any institution can use.
APIs as the New Operating System of Banking
If mainframes were the operating system of banking for 50 years, APIs are becoming the operating system for the next 50. An API (application programming interface) is a standardised way for one piece of software to communicate with another. In banking, APIs allow a bank’s core system to connect with payment networks, identity verification services, credit bureaus, fraud detection platforms, and customer-facing applications.
Banks globally process over 2 billion API calls daily, handling $676 billion in transaction value, according to Coinlaw. That volume reflects how deeply API-based architecture has penetrated banking operations. Every time a customer checks their balance, makes a payment, or receives a transaction notification, multiple API calls are firing in the background.
Open banking regulations in the EU (PSD2), UK, Australia, and Brazil have accelerated API adoption by requiring banks to share customer data through standardised interfaces. This regulatory push created a market for third-party applications built on bank data: budgeting tools, automated savings products, account aggregators, and alternative credit scoring models.
What the Data Shows About the Current Transition
The neobanking market, where cloud-native technology is the entire bank rather than a component added to a legacy system, reached $210.16 billion in 2025, per Fortune Business Insights, growing at 49.30% annually toward $7.66 trillion by 2034. Europe holds the largest regional share at 37.20%.
Cross-border payments, one of the banking functions most transformed by modern technology, reached $371.59 billion in 2025, according to Fortune Business Insights. Fintech companies using API-based connections to local payment networks have compressed settlement times from days to seconds and cut costs by 60% to 90% compared to traditional correspondent banking.
Traditional banks are responding with massive technology investment. JPMorgan Chase spent $17.1 billion on technology in 2024. But spending alone does not determine outcomes. The critical variable is architecture: whether a bank is adding features to a 40-year-old mainframe or building on a modern, API-native platform.
The evolution of banking technology is not a smooth continuum. It has moved through three distinct architectures, each replacing the one before. The mainframe era lasted 50 years. The channel-layer era lasted 20. The cloud-native era is five to ten years in, and the institutions that delay the transition are accumulating a technology debt that compounds with every passing quarter.
