Blog · Legacy systems
IBM lost about $31 billion on the stock market over one announcement. Translating COBOL with AI is only the first step of a migration
On February 23, 2026, Anthropic published how to use Claude Code to modernize COBOL systems, and IBM stock closed down nearly 13%, its worst day since October 2000. According to Forbes, the company lost about $31 billion in market value. But understanding the code is only the first part of a migration.
What happened
Anthropic explained on its blog that Claude Code can map dependencies across thousands of lines of code, document workflows nobody remembers, and identify risks that would take a team of analysts months to find. According to CNBC, that exploration and analysis work accounts for most of the complexity of modernizing COBOL, a key business for IBM.
That same day the stock fell nearly 13% and closed at $223.35, according to CNBC. Forbes calculated that market value dropped from $240.8 billion to about $208.7 billion. With that drop, the stock was down more than 24% for the year.
What the announcement itself says
Anthropic's post does not promise an automatic migration. It proposes moving component by component, validating each step, running the old and new code in parallel during the transition, and leaving the core decisions to people: business priorities, target architecture, coding standards, which test scenarios a domain expert validates by hand, and what performance is required.
That list is the hard part of any legacy system. Code can be read and rewritten; what's hard is moving data that has been piling up for decades, keeping the integrations with everything that depends on the system, and continuing to operate while it changes. AI makes the analysis cheaper. It doesn't replace the plan or the tests.