IBM’s longstanding dominance in enterprise computing faced a significant challenge on February 23, 2025, when its shares plummeted by as much as 13% in early trading. This dramatic decline wiped approximately $25 billion from the company’s market capitalization, following the announcement by Anthropic that its AI tool, Claude Code, could facilitate COBOL-to-modern-language migration. This capability threatens a lucrative segment of IBM’s consulting and software business, which has long relied on the maintenance and modernization of legacy COBOL systems.
COBOL, or Common Business-Oriented Language, was developed in 1959 and remains entrenched in systems that process an estimated $3 trillion in daily commerce. The language underpins about 95% of ATM transactions and 80% of all in-person financial transactions globally, as reported by Reuters. Despite continual predictions of its obsolescence, COBOL persists due to the high costs and risks associated with replacing it. Major banks and government agencies have experienced failed modernization projects, sometimes costing hundreds of millions of dollars, leading to the prevailing sentiment: if the existing code functions adequately, why change it?
Over the years, IBM has been the primary beneficiary of this cautious approach. The company profited from its mainframe hardware, the z/OS operating system, and consulting services, as businesses felt bound to their existing COBOL infrastructure. In 2023, IBM acquired the mainframe modernization startup Advanced Modernization and continued investing in its Watsonx AI platform, emphasizing the importance of this revenue stream to its overall strategy. Analysts at Morgan Stanley estimate that COBOL-related services and infrastructure contribute between $5 billion and $7 billion to IBM’s annual revenue, although the company does not disclose these figures explicitly.
Anthropic’s announcement outlined Claude Code’s capabilities, which enable developers to interact with its large language model directly from the command line. According to the company’s technical documentation, Claude Code can read, interpret, and translate COBOL programs into modern languages such as Java, Python, and Go, while preserving business logic and handling dependencies. Anthropic demonstrated this functionality with a sample COBOL banking application of approximately 500,000 lines of code, claiming completion of the initial translation in under 72 hours with “high fidelity” to the original business rules. A human review team subsequently spent two weeks verifying the output.
In reaction to the announcement, Wall Street’s response was swift and severe. Several factors contributed to the stock decline. IBM has spent the past three years marketing itself as an AI company, with CEO Arvind Krishna emphasizing that generative AI would drive future consulting demand. The irony of a rival’s AI product undermining one of IBM’s most stable revenue streams was not lost on investors. Additionally, a January 2025 survey by Gartner indicated that 42% of Chief Information Officers (CIOs) intended to reduce spending on legacy modernization consulting over the next two years, largely due to AI-assisted tools.
Anthropic’s announcement also validated a theory proposed by competitors such as Google and Microsoft, suggesting that large language models are particularly well-suited for understanding and translating legacy code. Google’s own code migration tools integrated within its Gemini platform have been gaining traction, although none have made as bold a claim regarding COBOL specifically.
In response, IBM moved to mitigate the fallout. The company issued a statement welcoming innovation in legacy modernization but cautioned that “translating code is only a fraction of the modernization challenge.” IBM highlighted its Watsonx Code Assistant for Z, launched in 2023, which aims to help clients refactor COBOL applications on its mainframe platform. An IBM spokesperson emphasized the importance of understanding decades of accumulated business logic and the integration points with numerous other systems, stating, “Anyone can translate syntax. The hard part is understanding the decades of accumulated business logic, the undocumented edge cases, and the integration points with hundreds of other systems.”
Despite IBM’s reassurances, skepticism remains prevalent. Industry insiders noted that the history of automated COBOL migration is fraught with challenges. Micro Focus, now part of OpenText, has offered COBOL migration tools for over two decades, achieving mixed results. The principal obstacle is not merely translating code but replicating the behavior of systems that have been customized over decades, often by programmers who are now retired or deceased, leaving behind little documentation.
“I’ve seen at least five waves of ‘COBOL is finally dead’ announcements in my career,” remarked a principal architect at a major U.S. bank, who spoke on the condition of anonymity. “Every time, the tools get better at the easy 80%. But the last 20% — the stuff that actually matters, the edge cases that handle billions of dollars — that’s where projects go to die.”
The implications for the broader IT services market are significant. The global IT services sector generates approximately $1.3 trillion annually, according to IDC, with a substantial portion allocated to maintaining and incrementally modernizing legacy systems. If AI tools can reduce modernization timelines by even 50%, the financial impact on firms like IBM, Accenture, Tata Consultancy Services, and Cognizant could be considerable.
Some analysts suggest that a positive scenario for IBM could emerge. If AI-assisted modernization indeed accelerates the migration away from mainframes, IBM might pivot to capturing revenue from cloud infrastructure and AI platform fees rather than traditional consulting hours. The company’s hybrid cloud strategy, centered around Red Hat OpenShift, is designed for this transition. However, the execution risks are substantial, as IBM would need to replace high-margin consulting revenue with typically lower-margin cloud fees in the short term.
Looking ahead, the competitive landscape is set to evolve. IBM is expected to enhance the capabilities of Watsonx Code Assistant for Z, with potential announcements at its upcoming Think conference. Both Google and Microsoft are likely to prioritize enterprise code migration within their AI strategies for 2025. Meanwhile, Anthropic, buoyed by significant funding from Amazon and other investors, has indicated that enterprise software development is a key component of its future roadmap.
For the thousands of enterprises still relying on COBOL, Anthropic’s announcement may represent a turning point, psychologically more than technically. The fact that a credible AI company claims it can manage COBOL migration is likely to encourage CIOs to revisit modernization proposals previously deemed too risky or costly. Whether Claude Code or any other tool can fulfill this promise at an enterprise scale remains uncertain, but it is now a question that every board with a mainframe budget will need to consider.
IBM’s stock partially rebounded in afternoon trading, closing down 8.4% for the day. The market’s message is clear: the protective moat surrounding IBM’s mainframe business is no longer viewed as impenetrable. In an era where AI can analyze and interpret legacy code at scale, the competitive advantages that once appeared permanent are being reassessed in real time.
