Anthropic’s claim that AI can quickly refactor COBOL rattles IBM investors

Anthropic’s assertion that its agentic coding tool Claude Code can rapidly refactor legacy COBOL code has unsettled IBM investors, triggering a sell-off in shares, which resulted in the stock registering a 13% drop on Monday — its biggest single-day drop since October 2000.

The assertion came in the form of a blog post from Anthropic that suggested that AI could take away the cost and complexity associated with modernizing a mainframe.

“AI excels at streamlining the tasks that once made COBOL modernization cost-prohibitive. With it, your team can focus on strategy, risk assessment, and business logic while AI automates the code analysis and implementation,” the company wrote.

Investors were concerned that Claude Code would erode the value of IBM’s strong and growing infrastructure services and consulting business, which depends on high-margin COBOL modernization on mainframes.

However, modernizing COBOL applications to counter the dwindling numbers of developers proficient in the programming language is not a novel idea.

IBM itself, in 2023, expanded the abilities of its generative AI-based Watson Code Assistant to include COBOL code translation into Java.

Hyperscalers, such as AWS, Microsoft, and Google, too, have been offering mainframe modernization tools, including AI services. IBM-spin off Kyndryl has also been partnering with hyperscalers to offer similar services.

Yet competition from hyperscalers and partners has so far failed to meaningfully slow IBM’s mainframe business.

During its fourth-quarter earnings calls last month, IBM CEO Arvind Krishna, in his prepared remarks, said that its Z family of mainframes had seen a 48% increase in yearly revenue — its highest in 20 years.

Krishna further pointed out that the increase in revenue from its Z line could be attributed to the efficiency of the Watson Code Assistant in modernizing COBOL applications, in turn making it easy for enterprise development teams to run the mainframes.

Moreover, IBM, in response to Anthropic’s assertions and to allay investors’ fears, put out a blog post of its own arguing that translating COBOL code is not equivalent to modernizing the underlying systems in a mainframe.

“Translation captures almost none of the actual complexity. The modernization challenge is not a COBOL language problem. It is everything the application runs on and integrates with,” Rob Thomas, SVP of IBM software, wrote in the post.