AI to help modernize COBOL systems: IBM stock plummets
After Anthropic introduced a tool for automating time-consuming COBOL modernization, IBM's stock plummeted over 13 percent.
(Image: LCV/Shutterstock.com)
Apparently because the AI Claude Code from Anthropic can take over particularly time-consuming processes when updating source code in COBOL, IBM's share price fell by more than 13 percent on Monday. In a blog post, the AI company explained that its technology can automate the lengthy exploration and analysis work that must be carried out before a migration. Thanks to AI, a COBOL code base can now be modernized within a period of quarters, whereas it previously took years. For IBM shares, this was the worst stock market day in more than 25 years, and in February alone, the company's stock has now lost more than 26 percent of its value.
No more “armies” of consultants needed
COBOL still runs on numerous mainframes, which are used in particular by banks in healthcare and retail due to their high reliability. Most of these mainframes are from IBM, and the US company generates a significant portion of its revenue from them. Modernizing such a system “previously required entire armies of consultants who spent years mapping workflows,” Anthropic now writes. As a result, only a few were willing to bear the high costs. Claude Code can now handle precisely these cost-intensive parts of a modernization, while humans must take over supervision and control.
Nevertheless, the extent of the sell-off remains surprising. IBM itself presented an AI assistant for the more efficient translation of COBOL into Java two and a half years ago to drive the modernization of mainframe applications. Only in the summer of 2025 did the IBM CEO state that the tool had found widespread use. Companies would use it to understand the millions of lines of source code in their systems and thus be able to decide what should be reasonably modernized. Speaking to Bloomberg, an analyst points out that mainframes could therefore be modernized more efficiently for a long time, but operators would not tackle it.
Videos by heise
COBOL (Common Business-Oriented Language) was developed in the 1950s and is rarely used today. At the same time, these are particularly important systems that cannot simply be replaced by newer alternatives. For companies, mainframes also mean immense “technical debt” that causes high costs. At the same time, they increasingly hinder access to data and the integration of modern technology. Furthermore, there are fewer and fewer developers for COBOL, which is why experienced teams must be retained for a lot of money. Investors in AI technology like Claude Code likely suspect a large potential for savings here.
(mho)
