Artificial Intelligence

IBM Plunges 13% as Anthropic Launches AI COBOL Tool

Do generic AI models fail at COBOL migration or can off-the-shelf tools handle this specialized work effectively?

FEB 24·Updated FEB 26

IBM Plunges 13% as Anthropic Launches AI COBOL Tool

Story

Image credit: Dominika Zarzycka/SOPA Images/LightRocket/Getty Images

The Facts

IBM's stock plummeted 13% on Monday, its worst single-day percentage loss since October 2000, losing around $31 billion in market value after Anthropic announced an AI coding tool to modernize COBOL, a decades-old programming language.
Forbes
Bloomberg
The Wall Street Journal
"Hundreds of billions of lines of COBOL run in production every day, powering critical systems in finance, airlines, and government. Despite that, the number of people who understand it shrinks every year," Anthropic wrote in a Monday blog post.
Anthropic
Business Insider
"Modernizing a COBOL system once required armies of consultants spending years mapping workflows," Anthropic added. "Tools like Claude Code can automate the exploration and analysis phases that consume most of the effort in Cobol modernization."
Anthropic
Bloomberg
COBOL handles an estimated 95% of ATM transactions in the United States, with hundreds of billions of lines of COBOL running in production every day, powering critical systems in finance, airlines and government.
Anthropic
CNBC
The Economic Times
IBM defended its mainframe business, stating that the value it delivers has nothing to do with COBOL and that, whether the application is written in COBOL, Java, or any other language, the platform provides the same guarantees.
The Register
IBM Newsroom
Bloomberg
IBM itself debuted a tool in 2023 that uses AI to work with COBOL, including converting it to Java, and CEO Arvind Krishna said in July 2025 that IBM's AI coding assistant for mainframes has seen very wide adoption among customers.
The Register
Ars Technica
Bloomberg

Sources Split

LEFT

CENTER

RIGHT

ESTAB
CRIT

CENTER

PRO
ESTAB

The Spin

Narrative A

Anthropic's latest tool directly threatens IBM's lucrative COBOL modernization services by automating the complex analysis and dependency mapping that once required expensive consultants and years of work. This development accelerates the erosion of IBM's mainframe dominance in legacy systems used by banks, insurers, and governments, justifying the sharp sell-off as investors recognize AI's potential to disrupt a key revenue stream.

Rui Diao on X

Narrative B

Generic AI models fail spectacularly at COBOL because they can't handle its unique structure, decades of embedded business logic, and the zero-tolerance error requirements that real migrations demand. Off-the-shelf tools like Opus 4.6 simply won't cut it for this specialized work. The market's panic over a single blog post completely ignores the actual technical reality that domain-specific training is absolutely essential for competent COBOL engineering.

Yang Li on X

Narrative C

Moving massive volumes of sensitive mainframe data to the cloud for AI processing introduces unacceptable latency and security risks, making the approach impractical. Data gravity and the need for real-time transaction processing mean AI must come to the mainframe, not the other way around. The platform has evolved into an AI-optimized hub with integrated accelerators that can run complex neural networks alongside mission-critical workloads.