# The surprising story of COBOL, the software that sustains digital capitalism

> Discover why COBOL, created in 1959, still sustains global banking and processes millions of transactions every day
> By José Maldonado · Technology · 2026-07-09

If at this very moment every server running COBOL in the world stopped working in unison, modern civilization would collapse before dawn. This is a very strong claim, perhaps exaggerated in the eyes of many, but the reality is that it is exactly what would happen.

Imagine it for a moment. You go to an ATM with your debit card and want to withdraw some cash, but it doesn't work. You want to make a card payment using a bank point-of-sale terminal, and that doesn't work either. You go to the bank to check your account, and there is no system and everything is chaos. Now imagine the situation you are experiencing with your money happening to everyone in the world at the same time, including digital payment platforms such as PayPal or Stripe, which have to stop operating because their banking connection is down globally.

And no, it is not a science fiction hyperbole, it is the reality of a programming language born in 1959 that, sixty-seven years later, still processes the financial heart of the planet. Because while technology trends are born and die in cycles of a few years, COBOL (Common Business-Oriented Language) remains unchanged, like a reinforced concrete colossus in the middle of a forest of glass skyscrapers, and no one has been able to dent its achievements.

But how did we get here? Well, that is the story we want to tell you here.

## The big gray box

In the late 1950s, computing was an archipelago of disconnected islands. Each computer manufacturer, with IBM at the helm, designed its own machines with its "own programming languages." That meant that if a company changed hardware, it had to rewrite all its software from scratch. Of course, this "programming" was more like weaving hundreds of cables to change the computer's logic and program it for a certain task. A slow, complex process that only a few chosen ones could do. The arrival of punched-card programming did not change that reality much, it only made it a little more practical.

![A look at a bank mainframe](https://content.observatorioblockchain.com/wp-content/uploads/2026/07/mainframe-bancario.webp)
***A look at a bank mainframe***

Faced with that situation, the Pentagon, tired of financing this inefficiency, convened a committee of experts in May 1959 at the Institute of the University of Pennsylvania: **the CODASYL committee (Conference on Data Systems Languages).** Among the brilliant minds that shaped this effort stood out US Navy Captain (and later Rear Admiral) [**Grace Murray Hopper**](https://es.wikipedia.org/wiki/Grace_Murray_Hopper).

Although she did not design COBOL alone, her philosophy was the cornerstone: **programming should not be reserved for pure mathematicians writing in cryptic symbols. It should approximate standard English.** It was she who gave rise to the written programming languages we see now in the digital world, with the creation of the A0 compiler in 1952 and later with B0 or FLOW-MATIC developed in 1955. By comparison, FORTRAN, developed by IBM, is the first high-level language used in production, and dates from 1957.

**[How does the Tax Agency investigate you? Spain shields the AI that selects taxpayers](https://observatorioblockchain.com/ia/espana-blinda-la-ia-que-selecciona-a-quien-investigar-en-hacienda/)**

## From FLOW-MATIC to COBOL

FLOW-MATIC was a high-level language by any measure. You wrote specific instructions in English, the compiler took this and translated it into machine language so that the computer would do its job. Here you have an example of a FLOW-MATIC program

![Example of code written in FLOW-MATIC, COBOL's predecessor](https://content.observatorioblockchain.com/wp-content/uploads/2026/07/flow-atic.png)
*Example of code written in FLOW-MATIC, COBOL's predecessor*

These 18 lines of code compare product inventories (list A and B) and divide them between those that have a price and those that do not. A basic function for maintaining a product warehouse, something that on a computer by usual means was complex to program, is reduced to 18 lines that anyone can read.

However, FLOW-MATIC had limitations, Hopper was aware of that and that is what she proposed to change in CODASYL, and she certainly succeeded. COBOL was born with a radically pragmatic purpose: to be a commercial, readable and portable language. Its syntax did not use complex mathematical abstractions, but verbs and narrative structures (ADD, SUBTRACT, MULTIPLY, MOVE, PERFORM). It was designed so that business executives could at least read the flow of the internal logic of their systems. The result was a verbose, direct and astonishingly efficient language for handling massive data files and exact-precision arithmetic operations.

## The evolution of COBOL and its battlefields

But what happened with this programming language? Why, being so relevant, does no one seem to use it now? Well, here there is a bit of myth and elitism. On the one hand, the language was consolidating itself as the commercial engine of the planet, with companies and especially banks adopting it, but on the other, formal academia declared war on it. The most famous and ruthless attack came from Edsger W. Dijkstra, a pioneer of computer science and Turing Award winner, who remarked:

> The use of COBOL cripples the mind; its teaching should, therefore, be regarded as a criminal offence.

For Dijkstra and the purists of structured programming, COBOL's verbose syntax, its initial inability to support recursion and, above all, its absolute dependence on the uncontrolled GO TO command (which jumped from one line to another breaking the logic of the code), resulted in unmanageable "spaghetti code." They considered that the language was inherently flawed and incapable of hosting clean mathematical thinking.

And here something must be highlighted: **COBOL is a Turing Incomplete language, that is, it cannot perform infinite computation**. Every COBOL program has a beginning and an end, a state defined algorithmically, which makes it very secure. Curiously, Bitcoin and its Bitcoin Script also follow this logic, applying a Turing Incomplete language.

![Example of code written in COBOL](https://content.observatorioblockchain.com/wp-content/uploads/2026/07/cobol-ejemplo.png)
*Example of code written in COBOL*

## Academic elitism

For academics this was a step backwards. Thinking in limited languages was absurd, they wanted advanced languages, they wanted to reach AI and advanced robotics, they wanted the world of the Jetsons, and COBOL was a stone in their shoe.

But, despite the academic jabs, the industry could not afford to abandon what already worked. So, instead of dying, COBOL responded by maturing. Over the decades, it assimilated each of the great software engineering reforms to patch its cracks and adapt to modern times. Yes, the COBOL of the 60s is not the same as the one we see now, although they remain compatible.

**[Free software: The invisible scaffolding of internet civilization](https://observatorioblockchain.com/en/technology/software-libre-el-andamio-invisible-de-internet/)**

## A linear evolution

And so we have this linear evolution:

*   **COBOL-68, the era of the Great Unification.** Here the first official standard arrived at the hands of ANSI. It consolidated the different variants that hardware manufacturers had created, achieving real portability between machines from different brands for the first time in the industry.
*   **COBOL-74, the arrival of structural tools.** It introduced the first concepts of modular design and improved file handling. The language was beginning to react to the theoretical currents of the time, although GO TO still dominated the landscape.
*   **COBOL-85, the end of Spaghetti Code.** The true structural response to Dijkstra's criticisms. It introduced explicit scope terminators (such as END-IF, END-READ, END-PERFORM), allowing logical structures to be nested cleanly and almost completely eliminating the need to use unconditional code jumps.
*   **COBOL 2002, the Leap to Objects.**  A radical paradigm shift. It broke its traditional purely procedural nature by adding native support for Object-Oriented Programming (OOP). From this standard onward, the old engine could encapsulate data and methods, facilitating its integration with modern libraries.
*   **COBOL 2014 and 2023, to the cloud and interoperability.** The modern revisions added native support for processing global data exchange formats such as XML and JSON. Current compilers from IBM, GCC or Micro Focus allow the code to interact with REST APIs, be deployed in Docker containers and run in public cloud environments (AWS, Azure) optimized for traditional x86 architectures.

## Free software above all

The most curious thing of all is that the COBOL ecosystem (compilers) is mostly free software. Except for IBM and Micro Focus implementations, the rest of the implementations are totally free. Even more curious, IBM and Micro Focus proprietary extensions are so well documented and accessible that free options such as GnuCOBOL and GCC COBOL (now merged projects) are capable of using them without problems.

**[The silent revolution: how ARM is redesigning the future of computing](https://observatorioblockchain.com/en/technology/como-arm-esta-redisenando-el-futuro-de-la-informatica/)**

In fact, GCC, the largest and most important free software compiler suite in the sector, [**has support**](https://www.devclass.com/development/2025/04/28/gnu-compiler-collection-151-released-cobol-support-improved-rust-compatibility-concerns/1618432) for COBOL 2002, 2014 and 2023, offering the greatest free software support for this language. This sends a powerful message: **COBOL is not dead, it is still at the heart of the financial world and it will not leave it.** 

## The veins of the global financial system

And this leads us to the next question **Why does banking prefer to pay astronomical salaries to senior developers or bring retired programmers back before shutting down its mainframes?** The answer lies in the scale of COBOL's design: it was conceived for exact decimal arithmetic. Unlike modern languages that use binary floating point (which can introduce subtle rounding errors in million-dollar financial transactions), COBOL processes numbers exactly as humans count money.

The following figures describe the impact of COBOL on the current financial sector:

*   95% is the percentage of credit card transactions globally that pass, at some point along their authorization route, through code written in COBOL.
*   80% of in-person commercial transactions in the world depend directly on this language.
*   At least 3 trillion dollars is the approximate volume of daily transactions that COBOL systems manage securely at a global level.
*   It is estimated that there are more than two hundred billion lines of active COBOL code in production in global critical infrastructure, exceeding the amount of code written in many of the fashionable languages combined for mission-critical systems.

That is the COBOL monster, a rival of more than 60 years, that languages like C/C++, JAVA, not even Rust, will be able to defeat.

And it is that COBOL is the living testimony of resilience in software engineering. It is not an obsolete language; it is a base infrastructure, the digital equivalent of the water pipes or the subway tunnels of a megacity. While the world celebrates the arrival of conversational artificial intelligences and ephemeral cloud architectures, in the planet's digital basement there is a PERFORM verb written half a century ago that has just been successfully executed a million times in the last second. And it will keep doing so tomorrow, without complaining, without failing.
