Code shock: Anthropic’s Cobol claim erases $30-billion from IBM’s market cap

Mumbai/Bengaluru: Anthropic’s claim that its AI coding tool Claude can understand and modernise programming language Cobol sent IBM shares down 13% on Monday, its steepest single-day fall in over 25 years, wiping out nearly $30 billion in market capitalisation.

Cobol is a decades-old programming language that underpins many systems running on IBM’s mainframes, which are used by critical infrastructure, including those of banks, airlines and govts. Around the world, there are very few Cobol developers anymore, the original developers have retired, and universities no longer teach the language. So, it’s increasingly becoming difficult to maintain these systems, and to rectify issues if something goes wrong. For that reason, converting the 800-billion lines of code in Cobol into more modern programming languages is seen as important.

But that currently is expensive to do. Anthropic’s contention is that Claude now makes it substantially more affordable.

Following the stock collapse, IBM’s senior vice president for software, Rob Thomas, responded that translating legacy Cobol code does not solve the broader modernisation challenge. “Translating code is one thing. Modernising a platform is something else entirely. The two are not the same, and the gap between them is where most enterprises run into trouble,” he said in a blogpost. Perhaps partly in response to this, the stock recovered almost 5% in early trade on Tuesday.

Thomas argued that Cobol on IBM mainframes has been optimised over decades of "tight coupling" with hardware that simple code translation cannot replicate. He compared it with the iPhone/iOS integration, and said, “someone could build an alternative (to the iPhone), but it is unlikely to displace a billion iPhones.” Thomas emphasised that the challenge lies not in Cobol itself, but in the wider ecosystem the applications run on and integrate with.

Asked about Anthropic’s announcement and its impact on legacy systems, HCLTech CEO C Vijayakumar said while technology is advancing rapidly, deploying it in complex enterprises is gradual. “The hype around instant modernisation often overlooks the real-world complexity of enterprise landscapes,” he said at the Nasscom Technology Leadership Forum (NTLF) in Mumbai. Indian IT firms have a significant book of business from banking and financial services.

Fractal Analytics founder & CEO Srikanth Velamakanni said if AI can bring down costs of code conversion, it would be a big opportunity for Indian IT. Currently, he said, migrating Cobol code is about $15-$20 per line. For a 50-million-line system, that could amount to $700 million to $1 billion and would take over a year. “The risk is high, complexity immense, and CIOs naturally avoid such high-stakes projects unless necessary. AI changes that dynamic. With AI-assisted modernisation, the cost per line could drop to about $2. That turns a $1-billion, 18-month project into $100 million,” he said, and noted that conversion of the entire 800 billion lines of code will be a $1.6 trillion opportunity for Indian IT.