IBM is the Latest AI Casualty: Shares Tank 13% on Anthropic's COBOL Modernization Threat
On Monday, February 23, 2026, Wall Street witnessed a shocking sight: IBM — one of the most storied names in technology — suffered its worst single day in over 25 years. Shares of International Business Machines Corp. (IBM) plummeted 13.2%, closing at $223.35 per share, wiping out nearly $40 billion in market value in a single session. The trigger? A blog post from AI startup Anthropic.
What Exactly Happened?
To understand this market shock, you first need to understand a programming language most people have never heard of: COBOL.
Short for Common Business-Oriented Language, COBOL was developed in the late 1950s. That’s not a typo — this language is older than colour television. Yet today, it quietly powers the backbone of global finance. According to Anthropic, 95% of all ATM transactions in the United States run on COBOL code. Hundreds of billions of lines of COBOL are running in production every single day, processing payments, managing bank accounts, handling airline reservations, and running government systems.
IBM has long been the undisputed king of COBOL infrastructure. The company sells expensive mainframe systems specifically optimised for high-volume COBOL-based transaction processing, and it charges clients millions to modernise these legacy systems. This is not a small side business — it is a core revenue pillar for IBM’s global consulting and technology services division.
On Monday morning, Anthropic published a blog post announcing that its Claude Code tool can now automate the most complex and expensive parts of COBOL modernisation. The market reacted with immediate panic.
Anthropic’s COBOL Play: What Claude Code Can Actually Do
Anthropic’s announcement was not vague marketing — it was a detailed technical playbook. The company claimed Claude Code can:
- Map dependencies across thousands — or even millions — of lines of legacy COBOL code
- Document workflows and business logic that were often never properly documented to begin with
- Identify risks that would previously take human analysts months to surface
- Simulate deployment scenarios to flag what is safe to migrate and what needs careful handling
- Reduce modernisation timelines from years to quarters
“Legacy code modernisation stalled for years because understanding legacy code cost more than rewriting it,” Anthropic wrote. “AI flips that equation.”
That last sentence is what rattled investors. IBM has charged billions over the years precisely because COBOL modernisation was so expensive, so complex, and so risky. If Claude Code can automate the most labour-intensive parts of that process, the entire economics of the business changes overnight.
Claude Code operates as an agentic AI system, meaning it does not just suggest code — it independently explores codebases, generates modernisation blueprints, and produces production-ready refactoring recommendations. This is qualitatively different from a simple AI chatbot. It behaves more like a senior consultant who never sleeps, never demands a raise, and can process millions of lines of code simultaneously.
IBM’s Worst Day Since the Dot-Com Crash
To put IBM’s 13.2% single-day decline into perspective: the last time IBM had a worse trading session was October 2000, when the dot-com bubble was bursting. IBM shares have now fallen 27% in February 2026 alone — on track for the company’s worst monthly performance since at least 1968, according to Bloomberg data.
That is a staggering collapse for a company that as recently as late 2025 was being positioned by analysts as a stable “enterprise AI pivot story,” thanks to CEO Arvind Krishna’s investments in the Watsonx AI platform and hybrid cloud infrastructure.
The speed of the selloff tells its own story. This was not a slow, measured reassessment by institutional investors carefully reading financial models. This was fear — a “sell first, ask questions later” market that has become increasingly common as AI disruption accelerates in 2025 and 2026.
IBM Pushes Back — With a Fair Point
To its credit, IBM did not stay silent. The company pushed back, arguing that Anthropic’s tool does not address the “fundamental engineering challenge of running mission-critical workloads at scale.”
And IBM has a point worth considering. Here is what the bears may be missing:
First, IBM has been in this space for years. In fact, the company launched its own AI-powered COBOL modernisation tool — Watsonx Code Assistant for Z — back in 2023. Anthropic’s blog post is not revealing a problem IBM was unaware of; it is threatening to out-execute IBM on IBM’s own turf.
Second, COBOL modernisation is not just a code translation problem. Large banks, insurance companies, and government agencies are not going to migrate their entire core banking infrastructure based on a blog post and a demo. These are systems where a single error can freeze millions of ATMs. The trust, regulatory compliance expertise, and enterprise relationships IBM has built over decades cannot be replaced by an AI tool overnight.
Third, Jefferies analysts noted in an investor brief that the threat may be less severe than the market is pricing in, and that IBM’s longer-term trajectory depends more on hybrid cloud and AI growth — areas where it remains competitive.
The Bigger Picture: AI’s Disruption Wave Hits Enterprise IT
IBM’s share collapse on February 23 was not an isolated event. It was the latest in a growing wave of AI-driven market disruptions:
- The week before, Anthropic unveiled Claude Code Security — an AI tool that can scan entire codebases for security vulnerabilities. That announcement sent cybersecurity stocks like CrowdStrike, Okta, Palo Alto Networks, and Cloudflare tumbling.
- The broader Dow Jones Industrial Average dropped 800 points on the same day IBM fell, as AI disruption fears compounded ongoing tariff concerns.
- Software stocks as a category lost over $200 billion in combined market capitalisation in the broader selloff.
What we are witnessing is a fundamental repricing of enterprise IT services companies. For decades, the complexity and opacity of legacy systems created a lucrative moat for companies like IBM, Accenture, and Cognizant. AI is threatening to make that moat crossable — not next decade, but now.
What This Means for Indian Banks and Global Finance
From my experience working in Indian banking for 15 years, this story has direct relevance beyond Wall Street.
Many large Indian public sector banks — including SBI, Bank of Baroda, and PNB — run core banking systems built on legacy technologies that share many characteristics with COBOL-era architecture. The global shift toward AI-driven legacy modernisation will eventually reach Indian shores, and the companies that have been charging premium consulting fees to manage these migrations should be watching closely.
For Indian investors, IBM is listed on US exchanges, and its Indian-listed peers in the IT services space — TCS, Infosys, and Wipro — all have significant legacy modernisation revenue streams of their own. Infosys chairman Nandan Nilekani said just last week that AI is making legacy rewriting “affordable and imperative.” If Claude Code’s capabilities are as advertised, the consulting fee structures across the entire enterprise IT services industry are under pressure.
Should You Buy, Sell, or Hold IBM?
This is not investment advice. Please consult a SEBI-registered investment advisor before making any financial decisions.
That said, here is the balanced view:
The bear case is real. If Anthropic’s Claude Code can genuinely compress COBOL modernisation timelines from years to quarters, IBM’s most defensible and profitable service lines face structural disruption. The $40 billion market cap erasure reflects a market that is taking this seriously.
The bull case is also real. IBM has deep enterprise relationships, regulatory credibility, and its own AI tools already deployed. Legacy modernisation at scale for tier-1 banks requires far more than code analysis — it requires programme management, regulatory sign-off, disaster recovery planning, and 24/7 support infrastructure. IBM provides all of this. An AI coding tool does not.
The most likely scenario: COBOL modernisation becomes cheaper and faster, IBM’s consulting margins compress, but the company is not made irrelevant. The real losers may be mid-tier IT service providers who lack IBM’s brand trust but also lack Anthropic’s AI capabilities.
Key Takeaways
- IBM shares fell 13.2% on February 23, 2026 — their worst single-day decline since October 2000
- The trigger was Anthropic’s announcement that Claude Code can automate COBOL legacy system modernisation
- COBOL powers 95% of US ATM transactions and remains critical to global banking infrastructure
- IBM has lost 27% of its value in February 2026 alone
- IBM is defending its position, arguing AI tools cannot replicate enterprise-scale engineering expertise
- Analysts at Jefferies believe the market reaction may be overdone
- The broader AI disruption wave is rapidly repricing enterprise IT services — a trend with global implications, including for India’s IT sector
Disclaimer: This article is for informational and educational purposes only. It does not constitute investment advice. The author has 15 years of banking experience and an MBA in Finance, but is not a SEBI-registered investment advisor. Always conduct your own due diligence before making investment decisions.
