IBM shares slid sharply after Anthropic claimed its AI can modernize COBOL systems. The selloff reveals deeper anxiety about legacy tech models in an AI-first world.
When International Business Machines shares tumbled after an announcement from Anthropic, it wasn’t because IBM missed earnings. It was because the market suddenly questioned something more structural.
Anthropic said its AI tools can help modernize COBOL code- the decades-old programming language that still runs core systems in banks, insurers, and governments. That might sound niche. It isn’t. COBOL modernization has long been slow, complex, and expensive. IBM has built a durable business around supporting and upgrading those legacy environments.
So when an AI firm suggests it can compress years of manual migration work into something far faster, investors don’t wait for proof. They react to the possibility.
IBM’s drop was sharp.
The scale of it says more about market psychology than immediate revenue risk. COBOL systems are deeply embedded. Enterprises don’t rip out mission-critical infrastructure overnight. AI can escalate parts of modernization. But oversight, compliance, and risk management still demand human involvement.
But here’s the nuance.
IBM’s strength has always been stability. Predictable enterprise contracts. Long-cycle infrastructure. Recurring services revenue. Anthropic’s pitch introduces uncertainty into that predictability. If AI tools reduce the labor intensity of modernization, margins in consulting and legacy support could tighten over time.
That doesn’t mean IBM is obsolete. It means the competitive terrain is shifting.
The real issue is perception. AI firms are now positioning themselves not just as product innovators, but as efficiency engines for legacy transformation. That reframes the value chain. Suddenly, AI isn’t just additive. It’s potentially deflationary for traditional service models.
IBM has navigated platform shifts before. Mainframes to services. Services for hybrid cloud. It understands reinvention. But the speed of AI iteration differs. Markets are pricing that speed, not today’s fundamentals.
This episode isn’t about COBOL alone. It’s about what happens when generative AI starts targeting the most entrenched corners of enterprise IT. Investors are asking a simple question: if AI can rewrite the past faster than consultants can bill for it, who captures the value?
Right now, the market isn’t sure IBM will.
