NEW YORK (TechGenez) – International Business Machines Corp. (IBM) shares suffered their steepest single-day decline in more than 25 years on Tuesday after AI startup Anthropic PBC unveiled capabilities in its Claude model designed to help modernize COBOL applications, a decades-old programming language that still powers a significant portion of IBM’s core mainframe business.
IBM stock fell as much as 18% during trading, its largest percentage drop since April 2000, before closing down 14.7% at $182.40, erasing roughly $35 billion in market capitalization in a single session. The sell-off was triggered by Anthropic’s demonstration that Claude Code can analyze, document, refactor, and translate legacy COBOL codebases into modern languages with high accuracy, directly addressing one of the most persistent technical debt challenges in enterprise IT.
Market Reaction and Trigger
The sharp move came after Anthropic released a video and technical paper showing Claude processing thousands of lines of production COBOL, generating accurate documentation, identifying business logic, suggesting optimizations, and producing equivalent code in Python, Java, or Go, tasks that traditionally require expensive, specialized consultants.
Wall Street analysts quickly revised earnings models, with several cutting 2026-2027 mainframe revenue forecasts by 8-15% and lowering price targets by 10-20%.
Morgan Stanley analyst Erik Woodring wrote: “Anthropic’s demo is a credible proof-of-concept that large language models can meaningfully accelerate COBOL modernization, an existential risk to IBM’s highest-margin segment.”
IBM’s Response
IBM moved swiftly to counter the narrative. In a statement released late Tuesday, the company said:
“IBM’s core mainframe platform, IBM zSystems, provides the same quality of performance, security, and reliability regardless of programming language. COBOL remains a small but important part of our mainframe ecosystem. The overwhelming majority of new workloads on zSystems are written in Java, Python, Go, Node.js, and other modern languages. We continue to see strong demand for hybrid cloud mainframe solutions that leverage AI to modernize applications in place without wholesale rewrites.”
IBM also pointed to recent mainframe client wins, including a multi-year deal with a major European bank and expanded adoption of IBM watsonx AI tools for mainframe modernization.
Broader Context
COBOL, developed in 1959, still runs an estimated 80% of global financial transactions and powers core systems at many of the world’s largest banks, insurers, and government agencies. IBM estimates that more than 220 billion lines of COBOL code remain in production worldwide, with annual maintenance costs in the tens of billions of dollars.
While mainframe revenue now accounts for roughly 20-25% of IBM’s total (down from over 40% two decades ago), it remains one of the highest-margin segments and a key differentiator in hybrid cloud deals.
Anthropic’s announcement follows similar tools from Microsoft (GitHub Copilot for COBOL), Google Cloud, and IBM’s own watsonx Code Assistant, indicating broad industry recognition that AI can accelerate legacy modernization.
Challenges
IBM faces multiple headwinds:
- Accelerating legacy modernization pressure from clients seeking cloud-native agility
- Growing competition in hybrid cloud from AWS, Azure, and Google Cloud
- Margin compression if mainframe revenue growth slows
- Need to prove watsonx can match or exceed third-party AI tools in COBOL refactoring
The company has invested heavily in AI for mainframe modernization, but analysts question whether it can move fast enough to offset the disruption risk.
Outlook
IBM is scheduled to report Q4 2025 earnings on January 29, 2026, where management is expected to address the Anthropic development and provide updated mainframe guidance.
Analysts will watch closely for signs of accelerated client migration away from legacy COBOL systems and any shift in watsonx adoption metrics.
While the sell-off was severe, some value investors view the drop as an overreaction, noting IBM’s diversified portfolio, strong free cash flow, and ongoing hybrid cloud momentum.
Conclusion
Anthropic’s demonstration that Claude can meaningfully accelerate COBOL modernization has delivered a sharp reality check to IBM investors, triggering the steepest single-day drop in more than a quarter-century. While the company insists its mainframe value proposition extends far beyond any single language, the event underscores how rapidly generative AI is reshaping perceptions of legacy technology risk and opportunity across enterprise IT.
