Investor confidence in IBM was shaken after Anthropic suggested that artificial intelligence could significantly simplify the modernization of legacy COBOL systems, raising fears about potential pressure on IBM’s high-margin mainframe services.
The concerns led to a sharp market reaction, with IBM shares falling 13 percent in a single trading session on Monday, marking the company’s steepest one-day decline since October 2000.
The market response followed a blog post published by Anthropic, which argued that its agent-based coding tool, Claude Code, is capable of rapidly refactoring legacy COBOL software.
The post maintained that AI-driven automation could substantially reduce the cost and complexity traditionally associated with mainframe modernization, allowing human teams to focus more on strategy, risk evaluation, and business logic rather than manual code analysis and implementation.
This claim unsettled investors who view IBM’s infrastructure services and consulting business as heavily reliant on the lucrative process of modernizing COBOL applications on mainframes. Concerns emerged that widespread adoption of AI tools such as Claude Code could weaken the value proposition of IBM’s long-standing modernization services.
However, efforts to modernize COBOL applications are not new. The decline in the number of developers skilled in the programming language has long driven enterprises to seek alternatives. IBM itself expanded the capabilities of its generative AI-powered Watson Code Assistant in 2023 to support the translation of COBOL code into Java.
Other major cloud providers, including Amazon Web Services, Microsoft, and Google, have also introduced mainframe modernization tools, many of which incorporate artificial intelligence.
In addition, Kyndryl, which was spun off from IBM, has partnered with hyperscalers to deliver comparable services. Despite this competitive landscape, IBM’s mainframe business has remained resilient.
During the company’s fourth-quarter earnings call last month, IBM’s chief executive officer Arvind Krishna reported that revenue from the Z family of mainframes rose 48 percent year on year, representing the strongest growth in two decades. The increase was attributed in part to the effectiveness of Watson Code Assistant in supporting COBOL modernization, which has helped enterprise development teams operate mainframes more efficiently.
In response to Anthropic’s claims and the resulting investor anxiety, IBM published its own blog post emphasizing that code translation alone does not constitute true mainframe modernization.
The company argued that the core challenge is the broader ecosystem of systems, infrastructure, and integrations that support mainframe applications, rather than the COBOL language itself.
IBM software leadership further stressed that while AI tools can assist with certain aspects of code conversion, they do not address the full complexity involved in modernizing mission-critical enterprise systems.
