Shares of IBM plunged to $239.96 (–3.4%) on March 24, extending a brutal selloff that began when Anthropic announced its AI coding tool could automate the painstaking work of modernizing COBOL — the ancient programming language that still powers most of the world's banking and government systems. IBM shares already suffered their worst day in more than 25 years on February 23, plunging 13% after Anthropic's initial announcement. The question for shareholders: is this a justified repricing or a panic driven by a blog post?

  • The Threat Hits IBM Where It's Weakest. If a general-purpose AI coding tool can automate COBOL modernization at a fraction of IBM's consulting rates, the addressable market for IBM's highest-margin services work shrinks.

Consulting was already IBM's slowest segment, growing just 3% year-over-year to $5.3 billion in Q3 2025. Anthropic's pitch — that AI can replace "armies of consultants" — targets exactly this vulnerability.

  • COBOL Runs Everything, and That's the Point. COBOL handles an estimated 95% of ATM transactions in the U.S., with hundreds of billions of lines running daily in finance, airlines, and government — yet the number of people who understand it shrinks every year. Anthropic claims its tool can compress modernization timelines from years to quarters. Modernizing a COBOL system once required armies of consultants spending years mapping workflows; AI tools can now automate the exploration and analysis phases that consume most of the effort.

  • IBM's Real Money Is Shifting Away From Consulting. Software, which accounted for nearly 45% of IBM's business, grew 11% year-over-year to almost $30 billion in 2025.

Full-year revenue hit $67.5 billion, up 8%, with net income jumping 76% to $10.6 billion.

IBM's mainframe hardware itself posted its highest annual revenue in about 20 years, powered by the z17 launch. The consulting arm investors fear losing is actually IBM's smallest growth engine.

  • Skeptics Say the Market Is Getting Ahead of Itself. Critics note the difference between a language model that produces "likely correct" outputs and the zero-error standard required when modernizing systems that clear billions of dollars annually.

Analysts caution IBM's mainframe footprint involves decades of dense integration across custom hardware, databases, and security models that no AI tool can simply replace. IBM itself debuted a COBOL-to-Java AI tool in 2023, and CEO Arvind Krishna said in mid-2025 it had achieved "very wide adoption."

The bottom line: IBM's stock has shed roughly $30 billion in market value, yet the threatened consulting segment accounts for under a third of revenue and is already its weakest link. With the stock trading around $240 versus a consensus analyst target near $325 — a 41% gap — the market is pricing in disruption that hasn't arrived yet.