Claude Code COBOL: How AI Wiped $40B Off IBM in One Day

On Monday, February 23, 2026, IBM had its worst trading day in over 25 years. Shares plummeted by 13.2%, closing at $223.35. The trigger? AI startup Anthropic published a blog post announcing that its Claude Code tool can now analyze, document, and modernize legacy COBOL codebases at unprecedented speed and scale. By the end of the session, investors had erased roughly $40 billion from IBM’s market capitalization. That marks the company’s steepest single-day percentage drop since October 2000.

The selloff reached far beyond IBM alone. It marks the third major AI-triggered stock collapse in just one week. Cybersecurity stocks had already dropped sharply on Friday after Anthropic unveiled Claude Code Security. The broader tech selloff also dragged crypto markets down, with Bitcoin pulling back to $64,000 the same day. AI capabilities are expanding faster than markets can absorb them.

Why COBOL Still Powers the Modern World

Developers created COBOL (Common Business-Oriented Language) in the late 1950s. They designed it as a human-readable programming language optimized for business transactions. It uses full decimal-point math, which made it ideal for financial systems, government operations, and enterprise computing.

More than six decades later, COBOL remains deeply embedded in modern infrastructure. According to industry estimates, roughly 95% of all ATM transactions in the United States still run on COBOL. The Open Mainframe Project puts the total at an estimated 250 billion lines of COBOL code in active production. These lines power critical systems across banking, insurance, airlines, and government agencies every single day.

The language itself works reliably and at massive scale. The challenge lies in the people who understand it. The engineers who originally wrote these systems are retiring. Efforts to train replacements have fallen short. This shrinking talent pool has turned COBOL modernization into one of the most expensive and risky undertakings in enterprise IT. IBM has built a significant portion of its revenue around solving exactly that problem.

What Claude Code COBOL Modernization Actually Does

On February 23, Anthropic published a blog post detailing how Claude Code can tackle COBOL modernization at scale. The company argued that AI can automate the exploration and analysis phases that historically consumed most of the effort and cost in legacy code projects.

According to Anthropic, Claude Code can map dependencies across thousands of lines of COBOL and trace execution paths. It can document workflows and identify risks that would traditionally take human analyst teams months to surface. Anthropic also released a Code Modernization Playbook alongside the announcement. The company pointed to existing YouTube demonstrations showing Claude Code working through live COBOL codebases.

The core argument was straightforward. Legacy code modernization has stalled for years because understanding the code cost more than rewriting it. AI flips that equation. Teams can focus on strategy, risk assessment, and business logic while AI handles code analysis and implementation.

Why IBM Stock Crashed After the Claude Code COBOL Announcement

IBM’s relationship with COBOL is structural. IBM manufactures most of the mainframe computers that run COBOL in production. The company generates significant revenue from both the hardware (its Z line of mainframes) and the surrounding ecosystem of consulting, modernization services, and infrastructure support.

Investors read Anthropic’s blog post as a direct threat to one of IBM’s most durable revenue streams. The logic was simple: if AI can dramatically cut the cost and complexity of COBOL modernization, IBM’s high-margin consulting engagements and mainframe lock-in could erode faster than expected.

The broader market environment amplified the reaction. Analysts describe the current investor mood as “sell first, ask questions later” when it comes to AI disruption. The pattern has repeated across multiple sectors in recent weeks. Cybersecurity stocks fell sharply after the Claude Code Security announcement [INTERNAL LINK: cybersecurity selloff article if available]. The COBOL news hit IBM with similar force.

With February’s losses, IBM shares have fallen approximately 27% in a single month. Bloomberg data shows the stock is on track for its worst monthly performance since at least 1968.

IBM Fires Back: Why COBOL Translation Falls Short

IBM pushed back quickly. In a blog post, Rob Thomas, SVP of IBM Software, argued that translating COBOL code covers only a small piece of the modernization puzzle, and the least complex part at that.

IBM pointed to the full stack of challenges beneath the code layer. Data architecture redesign, runtime replacement, transaction processing integrity, and decades of tight coupling between software and hardware define mainframe environments. Simply converting COBOL to Java or Python captures almost none of the actual engineering challenge, according to IBM.

IBM CEO Arvind Krishna noted in July 2025 that the company’s own AI coding assistant, Watson Code Assistant, had achieved wide adoption among mainframe customers. Most of them use it to understand their COBOL codebases and decide what to modernize. IBM also reported its highest mainframe revenue in 20 years in its most recent earnings. The company attributed part of that performance to its own AI modernization tools.

Analysts at Evercore ISI echoed IBM’s argument. Customers have long had the option to migrate away from mainframes, yet they consistently chose to stay. The platform’s determinism, reliability, and scalable compute remain unmatched for certain mission-critical workloads. Jefferies maintained a Buy rating on IBM with a $370 price target. The firm argued that the COBOL threat covers only a low single-digit percentage of overall revenue.

Can Claude Code COBOL Tools Actually Replace IBM’s Mainframe Business?

The opposing camp makes a compelling case as well. COBOL modernization has been a technically solvable problem for years. The prohibitive cost-to-benefit ratio kept it from happening at scale. If AI genuinely reduces the cost side of that equation, even incrementally, it could unlock a wave of migrations.

Matt Brasier, an analyst at Gartner, noted that modernization technology has existed for some time. Yet the costs remained high and the ROI remained low. If tools like Claude Code meaningfully shift that dynamic, the long-term implications for IBM’s mainframe business could outweigh a single day’s stock movement.

The competitive angle adds pressure too. Amazon, Google, and Kyndryl (an IBM spinoff) have all offered AI-powered COBOL migration tools in recent years. Claude Code adds another capable competitor to the mix. Anthropic’s rapidly growing reputation in the developer community gives it significant momentum. NAND Research chief analyst Steve McDowell put it clearly: Anthropic has a broader footprint within development teams. That reach makes them a worthwhile single-vendor option for enterprises already using their tools.

Skeptics on the AI side highlight a fundamental limitation, though. GenAI tools produce probabilistically correct outputs. The systems running COBOL demand 100% accuracy 100% of the time. Leading modernization tools combine deterministic and non-deterministic approaches. As Gartner’s Brasier noted, none of them fully solve the ROI problem on their own.

AI Disruption Is Repricing Legacy Tech in Real Time

The pattern here matters more than any single stock move. In one week, Anthropic’s announcements triggered massive selloffs across two entirely different sectors: cybersecurity and enterprise mainframe computing. Crypto markets followed, with Bitcoin and altcoins tracking the broader tech decline.

AI capabilities are expanding into established, high-margin business lines at a pace markets struggle to absorb. The repricing is happening in real time. Each new capability announcement from Anthropic, OpenAI, or Google serves as a stress test for incumbents. The results grow increasingly volatile.

What the Claude Code COBOL Selloff Means Going Forward

IBM will almost certainly weather this storm. The company’s mainframe business is deeply entrenched. Its customer relationships span decades. The technical barriers to full migration remain substantial. By Tuesday, the stock had already begun recovering, rising 2.67% as analysts broadly deemed the selloff overdone.

But the market’s reaction tells us something important: the era of “too complex to disrupt” is over. AI tools like Claude Code may not replace IBM’s mainframe ecosystem overnight. They are, however, lowering the barrier to entry for modernization. They are forcing a conversation that many enterprises have put off for years.

This is exactly where decentralized AI networks enter the picture. As centralized AI players like Anthropic reshape legacy industries, a parallel ecosystem of open, distributed AI infrastructure is emerging. This ecosystem could redefine how AI capabilities get built, owned, and deployed at scale. If you’re interested in how these trends converge, explore our coverage of the Bittensor ecosystem and what decentralized AI means for the future of technology and investment.