Original text here from Patrice Bernard (LinkedIn)
It took only a simple announcement from Anthropic to send IBM’s stock tumbling. The startup claimed that its AI assistant Claude Code could significantly accelerate the modernization of legacy systems written in COBOL—long considered one of the strongholds of IBM’s business. The market reaction was immediate: the company lost more than 13% of its value in a single trading session, its steepest daily decline in over two decades.
At first glance, the reasoning seems straightforward. For decades, the maintenance and modernization of COBOL systems have required large teams of specialized consultants working for months—or even years—to decipher undocumented codebases. If artificial intelligence can automate a large share of that effort, then the economic model underpinning those services might appear suddenly fragile.
Subscribe to our newsletter:
But the conclusion that IBM could become obsolete is based on a questionable chain of logic.
The idea that AI can assist with understanding and transforming legacy code is hardly new. Tools capable of analyzing dependencies, documenting workflows, and translating components into more modern languages have existed for years, and IBM itself has been investing heavily in the field. Its own generative-AI tools already perform tasks similar to those highlighted by Anthropic, such as identifying application dependencies or helping convert COBOL programs into Java.
More importantly, the real complexity of legacy modernization projects does not lie in reading the code. The hardest part is ensuring that new implementations reproduce exactly the same functional behavior as the original systems—often embedded in decades of undocumented business logic—and that they continue to interact correctly with surrounding infrastructure. These challenges involve governance, testing, and operational risk management that no AI can fully automate.
In other words, AI may accelerate certain technical tasks, but it does not eliminate the need for the broader expertise required to run mission-critical systems. And those systems remain pervasive: COBOL still processes a vast share of financial transactions and underpins numerous banking, government, and airline platforms around the world.
Seen from that perspective, the market’s reaction looks more like a symptom of the current AI hype cycle than a realistic assessment of IBM’s prospects. The company may face competition from new tools and vendors, but the structural demand for reliable infrastructure and modernization expertise is unlikely to vanish overnight.
The real lesson is not that IBM is doomed, but that the narrative surrounding artificial intelligence is powerful enough to reshape investor expectations—even when the underlying technological reality changes far more slowly.
