AI panic reels in America’s oldest tech giant

Matthew Field

Tue, February 24, 2026 at 11:40 PM GMT+9 5 min read

In this article:

IBM

+4.84%

ANTH.PVT

IBM, known as Big Blue, has become the latest US tech giant to be suffering from AI jitters - Stan Honda/AFP via Getty Images

Deep inside the server rooms of hundreds of the world’s biggest businesses, mainframe computers built by International Business Machines (IBM) hum and whirr.

IBM is ancient by US technology standards, founded more than a century ago. But its chunky computers form the backbone of banks, airlines, card payments and more.

Founded in 1911, it first manufactured early “punch-card” tabulation machines, before developing its mainframe technology in the 1960s. Some 70pc of Fortune 500 companies now rely on its machines, as do more than 90pc of the world’s banks.

Yet IBM, known as Big Blue, has become the latest US tech giant to be suffering from AI jitters. Its stock dropped 13pc on Monday, its worst day since 2000, on fears that a new tool from Anthropic could up-end its dominance of corporate IT.

The panic started with a blog. Anthropic, the Silicon Valley lab that develops the AI coding tool Claude, a rival to ChatGPT, claimed its technology could be used to dramatically cut the cost of maintaining and updating “Cobol”, a coding language developed in the 1960s that IBM’s machines are built on.

In particular, it said its AI could be used to help modern programmers make sense of Cobol and translate it into more up-to-date coding languages.

The archaic code is increasingly unfashionable and is rarely taught in computer science courses.

“Cobol is a very old language,” says Prof Alan Woodward, a computer science expert at the University of Surrey. “All the people that taught and wrote Cobol are of my era, and they are retiring or have retired. Consequently, it costs more and more to maintain it.”

With a dwindling pool of Cobol experts, businesses around the world rely on IBM to help keep these systems running. Critical industries, notably the financial sector, central banks, payments and even air traffic control systems, still rely on it. In the last quarter, IBM made $5.3bn (£3.9bn) from consulting fees, roughly a quarter of its revenues.

Rather than rely on aged experts from IBM, Anthropic suggested that AI could do the work.

“The economics of Cobol modernisation have shifted,” the company said. “AI makes the economics work by automating what used to require armies of consultants.”

Anthropic’s growing popularity

Despite only being founded in 2021, Anthropic already dwarfs IBM by value. The latest stock slide means IBM is worth around $200bn. Anthropic, on the other hand, was this month valued at $380bn.

As well as prompting a sell-off for IBM, shares in other IT consultants also dropped, with Accenture and Cognizant both falling more than 6pc.

Story Continues  

“I personally had the displeasure of coding in Cobol – and on IBM mainframe computers,” says Lukasz Olejnik, an independent technology consultant and visiting research fellow at the Department of War Studies at King’s College London.

Using AI to do this arduous work is an “obvious killer application”.

Anthropic’s Claude Code tool has proved popular with developers, since it can write and debug code semi-autonomously. The technology has become the tool of choice for so-called “vibe coding”, allowing engineers to work faster and those with little programming knowledge to code complex apps.

But Olejnik says the real benefit of Anthropic’s technology would come from speeding up the painful process of understanding just how a company’s sprawling Cobol codebase works, and how it interacts with the rest of the business, given the people who built it may have left years ago.

Previously, this would have taken IBM’s experts and consultants months. In its blog, Anthropic said this was akin to “reverse engineering business logic from systems built when Nixon was president”.

“AI can automate the analysis phase, that has always been the expensive part,” says Olejnik.

The sharp plunge in IBM shares comes as fears of an AI bubble grips the stock market. Investors are rushing to sell off their positions in companies seen as at risk from Silicon Valley’s new generation of AI labs. Whether they actually are vulnerable remains to be seen, but the market is in a “sell first, ask questions later” mood.

A memo from Citrini Research headed “The 2028 Global Intelligence Crisis” was blamed on Monday for a mass sell-off of some of America’s biggest tech names.

The note itself was part stock market research, part speculative fiction. It argued that AI would cause mass job losses that trigger an economic crash. It claimed that US unemployment would hit 10pc by 2028 and that the stock market would peak in October 2026, before collapsing.

The note named companies such as Deliveroo-owner Doordash and American Express as potential victims of AI.

Researchers claimed people could vibe code their own version of Doordash and argued that American Express would be undermined by people shopping through chatbots.

Citrini claimed that the technology led to a surge in “Ghost GDP” that benefited AI companies, but not the wider economy. “For every new role AI created, it rendered dozens obsolete,” the research stated.

The research was “forwarded to me around 10 times late last week”, wrote Jim Reid, an analyst at Deutsche Bank on Tuesday. Despite leaning “heavily on narrative and emotion rather than hard evidence”, it appeared to have triggered a rout across US software companies.

Despite the collapse in its stock, IBM has insisted that its business remains solid. Rob Thomas, its chief commercial officer, wrote on Monday: “The value IBM mainframe delivers has nothing to do with Cobol.

“Translating code is one thing. Modernising a platform is something else entirely.”

IBM has its own coding bot WatsonX that is designed to help solve the problem of updating mainframes that Anthropic is targeting. Its quarterly sales, meanwhile, climbed 12pc to $19.7bn in the final three months of 2025.

Woodward, of the University of Surrey, adds that critical businesses, such as banks, are unlikely to unleash an AI bot on their computers, given the financial risks involved.

“These systems are not trivial. Someone in charge of these systems is not going to let AI loose on it,” he says.

Even so, the IBM sell-off shows no one is immune to the AI panic.

Try full access to The Telegraph free today. Unlock their award-winning website and essential news app, plus useful tools and expert guides for your money, health and holidays.

Terms and Privacy Policy

Privacy Dashboard

More Info