IBM shares suffered their worst single-day drop in over 25 years on February 23, 2026, plunging as much as 13.2% and closing down 13.1% at roughly $223.35. The sell-off erased approximately $31 billion in market value in a single session and pushed the stock down 27% for the month so far—on track for its weakest February since the 1960s.
The trigger was a single blog post from Anthropic. The company introduced Claude Code, an AI capability designed to supercharge the modernization of legacy COBOL systems. COBOL, a 65-year-old programming language, still runs the world’s most critical infrastructure: 95% of U.S. ATM transactions, Social Security payments, major bank batch processing, airline reservations, and countless government back-ends. Hundreds of billions of lines of COBOL code sit on IBM mainframes.
For decades, modernizing these systems has been an enormously profitable business for IBM Global Services and its partners. Projects routinely took years and involved large teams of expensive consultants to untangle undocumented, sprawling codebases. That complexity created a durable economic moat of high-margin services and sticky hardware revenue.
Anthropic claims Claude Code changes the game by automating the most painful early stages: reading entire massive codebases, mapping thousands of interdependencies, tracing data flows, generating architecture diagrams, risk reports, and even test scaffolding. The result, according to Anthropic: “Teams can modernize their COBOL codebase in quarters instead of years.”
Wall Street heard one clear message: AI is about to eat IBM’s legacy cash cow. The broader software and IT-services sector already under pressure from AI disruption fears saw the same sharp moves. Accenture and Cognizant shares also fell sharply.
Yet some analysts called the reaction overdone. IBM has offered its own watsonx Code Assistant for mainframe modernization since 2023. Mainframe revenue actually reached a 20-year high in Q4 2025. Full production migration still demands rigorous human oversight, regulatory approvals, and extensive testing—especially for systems that literally cannot fail.
Still, the market’s verdict was swift and brutal. Investors are increasingly pricing in a future where AI coding agents dramatically shrink multi-year consulting budgets across the industry.
For IBM, today’s drop is a loud warning shot. The company’s long-term success may now hinge less on protecting its 1960s-era fortress and more on how quickly it can turn AI into an ally rather than an existential threat.