It’s an open secret (that’s, not many individuals appear to know) that the establishments protecting the worldwide monetary system turnig over run code that’s historic, barely understood, and frighteningly exhausting to exchange. Now, AI is lastly making that drawback solvable – and the market has responded with a actuality verify for one among expertise’s oldest names.
IBM shares recorded their worst single-day drop in additional than 25 years earlier this week, plunging 13% after AI startup Anthropic stated its Claude Code instrument can speed up COBOL modernisation – the sort of painstaking, costly legacy work that has underpinned a portion of IBM’s consulting income for years.
An Anthropic weblog stated that “modernising a COBOL system as soon as required armies of consultants spending years mapping workflows,” and argued that instruments like Claude Code can now automate the exploration and evaluation phases that eat many of the effort in COBOL modernisation. That single declare was sufficient to ship traders reaching for the promote button.
COBOL is greater than most realise
To know why the response was so sharp, it helps to grasp simply how entrenched COBOL stays. A whole lot of billions of traces of COBOL code run in manufacturing day by day, powering essential methods in finance and authorities sectors. The language handles an estimated 95% of ATM transactions within the US alone.
The deeper drawback isn’t the code itself – it’s the individuals who perceive it. The variety of builders who perceive COBOL continues to shrink because the workforce that constructed these methods has largely retired. That expertise shortage is exactly what made COBOL modernisation so costly for therefore lengthy, and what made massive consulting engagements – the sort IBM and rivals like Accenture and Cognizant constructed worthwhile practices round – basically unavoidable.
Anthropic argues that AI flips this equation totally. Claude Code works by mapping dependencies in hundreds of traces of code, documenting workflows, figuring out dangers sooner than human analysts, and offering groups with deep insights for knowledgeable decision-making. The corporate says groups can now modernise COBOL codebases in quarters not years.
IBM was already right here
What the market’s response could also be overlooking is that IBM itself has been making this argument for a while. Anthropic’s put up comes about three years after IBM itself suggested utilizing AI to rewrite COBOL as Java and created a product known as “watsonx Code Assistant for Z” to do it. IBM CEO Arvind Krishna stated as just lately as July 2025 that the corporate’s AI coding assistant for mainframes “has received very adoption,” with nearly all of prospects utilizing it to grasp their COBOL codebase and resolve what to modernise.
IBM defended its place on Monday, saying its mainframe platform delivers the identical high quality of efficiency and safety no matter programming language – COBOL or in any other case. And analysts had been fast so as to add nuance to the panic.
Evercore ISI analyst Amit Daryanani famous that “purchasers already had the choice emigrate from the mainframe, but they’re sticking with the platform,” suggesting the concern of displacement could also be outrunning the fact.
The broader sample
IBM wasn’t alone in taking a success. Accenture and Cognizant additionally declined following the information – an indication that traders are trying on the total consulting mannequin round legacy modernisation, not IBM’s mainframe {hardware} enterprise. Simply final week, cybersecurity shares bought off sharply after Anthropic introduced Claude Code Safety, a instrument that scans codebases for vulnerabilities.
The sample is turning into acquainted: every new AI capacity announcement triggers a reassessment of which current income streams is likely to be compressed, and the market costs in concern instantly.
IBM didn’t keep quiet. Rob Thomas, the corporate’s Senior Vice President and Chief Industrial Officer, pushed again instantly within the aforementioned blog post, drawing a line the market appeared to have missed: “Translating code is one factor. Modernising a platform is one thing else totally. The 2 should not the identical, and the hole between them is the place most enterprises run into bother.”
His argument is price sitting with. The worth IBM’s mainframe delivers, Thomas contends, has nothing to do with COBOL as a language – it lives within the vertically built-in stack beneath it: z/OS, transaction processing structure, quantum-safe encryption, and a long time of hardware-software optimisation that no code translation instrument touches.
Anthropic’s Claude Code, in his studying, is fixing an actual drawback – simply not the one which issues most for enterprises operating IBM Z. He additionally raised a degree that complicates the headline narrative additional: roughly 40% of COBOL truly runs on Home windows, Linux, and different distributed platforms – not mainframes in any respect.
A lot of what’s being framed as an IBM mainframe story is partly a distributed methods drawback that has been folded right into a mainframe headline. IBM’s personal purchasers are already making the case.
Royal Financial institution of Canada has used IBM’s watsonx Code Assistant for Z to map dependencies and construct modernisation blueprints for core purposes. The Nationwide Organisation for Social Insurance coverage reported a 94% discount in time to analyse legacy COBOL code utilizing the identical instrument – chopping an eight-hour process to roughly half-hour.
Whether or not Monday’s selloff was a good verdict or a reflexive one, the underlying change is actual: AI is making COBOL modernisation economically viable for the primary time in a long time. The query IBM is asking – and the market hasn’t absolutely answered – is whether or not that’s a menace to its enterprise or an acceleration of the transformation it’s already main.
See additionally: Hitachi bets on industrial experience to win the bodily AI race
Wish to study extra about AI and massive information from trade leaders? Take a look at AI & Big Data Expo going down in Amsterdam, California, and London. The great occasion is a part of TechEx and co-located with different main expertise occasions. Click on here for extra data.
AI Information is powered by TechForge Media. Discover different upcoming enterprise expertise occasions and webinars here.
