AI wipes $31 billion off IBM's market value over COBOL coding

In a single trading day, IBM lost market value comparable to the capitalization of a number of large tech companies. The stock plunged, erasing $31 billion from brokerage accounts, the worst result in almost three decades. The trigger for the panic was the seemingly archaic programming language COBOL, and the quiet but devastating ability of modern AI models to work with it as confidently as a programmer with thirty years of experience.

To grasp the scale of what happened, some context helps. COBOL, a language created back in 1959, is still the backbone of the world's financial infrastructure. By various estimates, systems written in it process more than three trillion dollars in financial transactions every day: banks, insurers, government agencies, pension funds. The problem is that COBOL specialists are catastrophically scarce. Their average age passed sixty long ago, and universities stopped training such programmers in the last century. IBM built its business for many years in precisely this gap between outdated technology and a shortage of skilled people: expensive consulting, multi-year contracts to maintain legacy systems, and manual code migration all brought in a stable and very substantial income.

Now that model is coming apart at the seams. Generative AI models have learned to handle COBOL with an accuracy that industry specialists considered unattainable not long ago. Modern systems can not only read and explain COBOL code but also translate it into modern programming languages such as Java, Python and Go, while preserving business logic that is sometimes documented nowhere and exists only in the code itself. What once took hundreds of engineers years of work and hundreds of millions of dollars in budget can now be done in a fundamentally different way: faster, cheaper, and without dependence on the exclusive expertise of a single vendor. Investors grasped this instantly.

The market's reaction exposed a deep contradiction in IBM's position. The company itself invested heavily in AI and publicly declared a technological transformation, yet investors suddenly realized that this transformation hits IBM first of all. If clients can handle migration and modernization of legacy systems on their own, or with the help of cheaper AI tools, why pay IBM millions for the same thing? The stock plunged, recording its worst one-day drop since 1998. This is not the panic of an irrational crowd. It is a sober recalculation of future cash flows for a company whose business model rested on technological complexity that AI is now starting to dissolve.

The consequences reach far beyond a single company. The IBM story is a vivid demonstration that AI disrupts not only low-skilled labor but also highly skilled niche expertise, which has traditionally commanded huge fees precisely because it was rare and hard to replicate. The enterprise IT consulting market, one of the most stable and highest-margin segments of the technology industry, has for the first time faced a direct and measurable threat from generative models. Analysts have already begun revising their estimates not only for IBM but also for other large consulting firms with a similar dependence on the complexity of legacy systems.

It is also telling that the blow came from COBOL, a technology that for decades was called "too old for automation" and "too specific for AI." This undercuts one of the main arguments for the resilience of traditional IT consulting: that corporate legacy systems are so unique and tangled that no automation can handle them. It turns out they can be automated. And that changes not only IBM's position but the entire pricing logic of the enterprise IT transformation market.

IBM now has to rethink what exactly it sells to clients in a world where the barrier to COBOL expertise has dropped sharply. A company with a history of a century and a half has reinvented itself more than once: from tabulating machines to mainframes, from hardware to services, from services to the cloud. The question is whether it can do so fast enough before the market reprices it for good.