IBM Just Had Its Worst Day in 25 Years. The Reason Is Hiding in Your ATM.

IBM shares fell 13.15% on Monday (their biggest single-day drop since the dot-com crash in 2000) after AI company Anthropic published a blog post announcing that its Claude Code tool can automate COBOL modernization. If that sentence sounds like a foreign language, don't worry. By the end of this, it'll make total sense.

What Is COBOL?

COBOL stands for Common Business-Oriented Language. It's a programming language invented in 1959, the same year Hawaii became a state, and COBOL never really went away. About 95% of ATM transactions in the United States still run on COBOL. Your bank's back-end system very likely uses it. So do most airline reservation systems, government payroll systems, and Medicare payment processing.

The reason it never went away isn't because it's great software. COBOL stuck around because replacing it is terrifyingly expensive and risky. These systems have been modified and patched for 60+ years. They contain decades of institutional knowledge, often with incomplete documentation. Many companies have tried to replace their COBOL systems and failed spectacularly or abandoned the project halfway through after running out of money.

Where IBM Comes In

IBM built a massive, profitable business around this problem. When companies need to maintain, upgrade, or modernize their COBOL systems, they often turn to IBM's consulting arm and mainframe infrastructure. IBM sells not just the hardware these systems run on, but the expertise to keep them running and, eventually, help migrate them to something newer.

This is unglamorous work, but it's lucrative and sticky. Once a major bank or government agency is deeply embedded in IBM's ecosystem, switching is expensive and painful. IBM's own CEO pointed to mainframe business strength as recently as last month, describing a 3x to 4x revenue multiplier from services and support contracts attached to mainframe placements.

What Anthropic's Announcement Changes

Anthropic's blog post laid out a specific claim: that the most expensive and time-consuming part of COBOL modernization can now be largely automated by AI. This include mapping what the code actually does, tracing dependencies across thousands of lines, documenting hidden workflows, and identifying risks.

The pitch is direct: what once required "armies of consultants spending years mapping workflows" can now be done by Claude Code in a fraction of the time. The company released a COBOL Modernization Playbook alongside the announcement and said AI can help teams modernize codebases "in quarters instead of years."

If true, this changes the economics that have kept companies locked into IBM's mainframe consulting ecosystem. The moat isn't completely gone. IBM still has relationships, certifications, and deep integration with critical infrastructure, but the case for paying premium consulting rates just got harder to make.

How Did the Market React?

IBM dropped 13.15%, wiping out roughly $30 billion in market cap in a single session. IBM is now down ~27% for the month of February alone, which is its worst monthly performance in decades. The sell-off also dragged down Accenture and Cognizant, which do similar legacy systems consulting work.

It's worth noting some skepticism here. IBM has been dealing with COBOL modernization pressure for years and has its own AI tools (watsonx) targeting the same space. And actually migrating a major bank's core systems is a multi-year, high-stakes project that goes far beyond code analysis. The process involves regulatory approvals, parallel testing, disaster recovery planning, and organizational change management. AI can help with the analysis phase, but it doesn't eliminate the need for human expertise in execution.

The Bigger Pattern

IBM is the latest in a series of companies hit by what investors are calling the "AI scare trade": the rapid repricing of businesses whose competitive advantage is built around expensive human expertise. In recent weeks, AI disruption fears have hit cybersecurity companies (after Anthropic launched a security scanning tool), commercial real estate firms, logistics companies, and financial services. The pattern is consistent: Anthropic or another AI company announces a new capability, and investors immediately ask which incumbent business model it threatens.

Whether the fear is justified in each case is a separate question. What's clear is that markets are increasingly treating AI disruption as a real and present risk.

Sources: CNBC, Bloomberg, Anthropic, Yahoo Finance, IBM earnings