IBM has survived the mainframe era, the PC revolution, the cloud transition, and a dozen industry upheavals over its 113-year history. But on February 23, 2026, a single AI announcement sent its stock into its steepest single-day freefall since the dot-com collapse — and forced a genuinely difficult question onto Wall Street's agenda: Is IBM's deepest competitive moat quietly evaporating?

Shares of IBM closed down 13.2% at $223.35, erasing more than $31 billion in market capitalization in a single session. It was the company's worst day since October 2000. And the catalyst wasn't a missed earnings report or a regulatory probe. It was a blog post from Anthropic.

What Anthropic Actually Said

Anthropic, the AI safety company behind the Claude family of models, announced that its Claude Code tool is now capable of accelerating COBOL modernization — specifically, the painstaking exploratory and analysis work that has historically been the most time-consuming (and expensive) phase of legacy system overhauls.

According to Anthropic, Claude Code can map dependencies across thousands of lines of code, document business logic, and identify architectural risks that "would take human analysts months to surface." For context, COBOL — the Common Business-Oriented Language first developed in the late 1950s — still powers roughly 95% of ATM transactions in the United States and underpins hundreds of billions of lines of active code at banks, insurers, and government agencies worldwide.

The implication was sharp and immediate: if AI can compress months of billable consulting hours into days of automated analysis, the economics of COBOL modernization contracts — a multi-billion-dollar business IBM has long dominated — could shift dramatically.

Why the Market Panicked

To understand the severity of the reaction, you need to understand what IBM actually sells in 2026.

IBM's consulting and infrastructure services divisions have been the company's crown jewels in its post-hardware transformation. Modernizing legacy mainframe environments — migrating COBOL workloads, redesigning data architectures, rebuilding transaction processing systems — is slow, complex, highly specialized work. The complexity is the business model. It creates long-term client relationships, recurring revenue streams, and formidable switching costs.

Claude Code, in theory, threatens to commoditize precisely that complexity. And investors, already skittish after watching AI disrupt SaaS incumbents like Salesforce, Adobe, and ServiceNow in recent months, didn't wait around to find out if the threat was real.

IBM's February selloff now stands at 27% — on track to be the company's worst month in over 50 years.

IBM Pushes Back — And They Have a Point

IBM did not stay quiet. Senior Vice President Rob Thomas issued a pointed rebuttal: "The value IBM mainframe delivers has nothing to do with COBOL. Whether the application is written in COBOL, Java, or any other language, the platform provides the same guarantees. The language is not the source of that value."

It's a fair argument, and one that deserves more credit than the market gave it on Monday. COBOL migration is not simply a translation problem. The real challenges — data architecture redesign, runtime replacement, transaction processing integrity, and hardware-accelerated performance built over decades — are not solved by better code analysis tools. They require domain expertise, regulatory knowledge, and organizational change management that no AI tool is close to automating away.

IBM also added, almost pointedly, that "new AI tools emerge every week," suggesting the company views the Anthropic announcement as one competitive data point among many.

And the fundamentals back them up: just last month, IBM reported its highest mainframe revenue in 20 years, with CEO Arvind Krishna attributing the strength in part — ironically — to AI code conversion tools driving morecustomer engagement with the platform, not less.

The Analyst Community: Measured, Not Alarmed

Wall Street's sell-side was notably more composed than the trading floor.

Evercore ISI maintained its "Outperform" rating and $345 price target, noting that IBM's latest z17 mainframe cycle is outperforming the prior z16 cycle across its first three quarters. Analyst Amit Daryanani wrote: "While we understand why mainframe migration could be a perceived negative for IBM, we would point out that IBM has already provided customers with several modernization options."

Jefferies echoed similar skepticism about the selloff's proportionality, arguing that IBM's long-term trajectory is more meaningfully tied to growth in hybrid cloud, AI, and data infrastructure — areas where IBM has been quietly building with its watsonx platform — rather than COBOL migration contracts alone.

The consensus view: the threat is real but likely overstated in the short term, and the stock's reaction may have been a case of pattern-matching fear (AI disrupts incumbents) overwhelming fundamental analysis.

The Bigger Picture: AI as Structural Risk, Not Just Market Noise

Here's what gets lost in the day-to-day price action: Anthropic's announcement is less a declaration of war on IBM and more a leading indicator of a broader structural shift that every enterprise IT incumbent needs to reckon with.

The consulting model that generates billions in revenue for IBM — and Accenture, and Cognizant, and dozens of systems integrators — rests on information asymmetry. Clients pay for expertise because the complexity of their own systems is opaque to them. AI tools that make that complexity transparent and navigable erode that asymmetry. Not overnight. Not completely. But directionally and persistently.

IBM's best answer to this is not to argue that the tools are limited (even if they are today). It's to be the company building and deploying those tools on behalf of enterprise clients — embedding AI into its consulting workflows, not defending against it. There are signs IBM understands this, but execution at the pace AI is moving remains the open question.

Buy the Dip or Brace for More?

At $223 per share, IBM trades at a meaningful discount to analyst consensus targets. The Evercore $345 target implies roughly 55% upside from current levels. The mainframe business, despite the narrative, just posted record revenues. And IBM's AI consulting pipeline — powered by watsonx and its growing ecosystem of enterprise AI partnerships — is not nothing.

But the market is also pricing in something real: the possibility that the comfortable, high-margin, slow-to-commoditize consulting business that IBM rebuilt itself around is entering a period of structural pressure it cannot fully control.

IBM has reinvented itself before. The question in 2026 is whether it can reinvent itself fast enough — and whether Claude Code just made the clock on that reinvention tick a little louder.

Sources & Further Reading