Shares of International Business Machines plunged 13% on Monday, February 23, 2026, after Anthropic announced that its artificial intelligence technology can modernize COBOL, the decades-old programming language that still powers much of the world’s banking and government infrastructure. The single-day loss was IBM’s steepest since 2000, rattling investors who had viewed the company’s legacy mainframe business as a durable revenue stream. The selloff struck at the heart of a tension that has been building for years, whether AI tools can displace the highly specialized, highly profitable consulting work that keeps aging enterprise systems running.
IBM’s Worst Trading Day in 25 Years
The 13% decline represented IBM’s worst single session in a quarter century, a distinction that places the drop alongside the dot-com era collapses that reshaped investor expectations for legacy tech firms. The trigger was specific: Anthropic said its AI could handle the tedious, expensive work of converting COBOL codebases to modern languages, a process that enterprises have historically outsourced to armies of specialized programmers. That claim cut directly at a revenue line IBM has defended for decades.
What made the market reaction so severe was the speed at which investors repriced IBM’s exposure. COBOL maintenance is not a side business for IBM; it is woven into the company’s software, consulting, and infrastructure segments. The company’s 2024 annual report filed with the SEC explicitly flags competitive risks from emerging technologies across those divisions. When Anthropic offered a concrete product that could automate a chunk of that work, the market treated it as an existential pricing event rather than a distant hypothetical.
Why COBOL Still Matters to Big Tech Revenue
COBOL, originally developed in the late 1950s, remains embedded in critical transaction-processing systems at major banks, insurance companies, and federal agencies. The language is notoriously difficult to update, and the pool of programmers who understand it has been shrinking for years as they retire. That scarcity has been a profit engine for firms like IBM, which charge premium rates for modernization consulting and ongoing support contracts. Anthropic’s claim that AI can now handle this conversion work threatens to compress what has been a high-margin, low-competition business.
The dynamic is straightforward: if an AI tool can translate COBOL to Java or Python at a fraction of the cost and time, enterprises have far less reason to pay for human-led modernization projects. IBM has been investing heavily in its own AI and hybrid cloud offerings, but those products are designed to complement its existing services business, not replace it. An outside competitor offering to automate the replacement itself poses a different kind of challenge, one that attacks the revenue base rather than adding to it. Reuters reported the selloff came directly after Anthropic’s announcement, confirming the market drew a straight line between the AI startup’s product and IBM’s bottom line.
The Market’s AI Disruption Calculus
Wall Street’s reaction reflects a broader shift in how investors evaluate tech incumbents against AI-native startups. For years, IBM’s mainframe and consulting businesses were seen as defensive assets, generating steady cash flow precisely because the underlying technology was so old and so difficult to change. Anthropic’s announcement inverted that logic. The same complexity that kept clients locked in now looks like a vulnerability, because AI tools are increasingly capable of handling tasks that once required deep institutional knowledge. The coverage from Bloomberg on the drop framed it as a direct consequence of Anthropic promoting COBOL modernization capabilities, underscoring how quickly sentiment can shift when a credible automation threat emerges.
One assumption worth questioning in the current coverage is that Anthropic’s tool will work as advertised at enterprise scale. Converting COBOL is not simply a translation problem. Legacy systems often contain decades of undocumented business logic, edge cases, and interdependencies that even experienced human programmers struggle to untangle. No public benchmarks or independent audits of Anthropic’s coding AI have been released, and the gap between a product demo and reliable production-grade migration is significant. Investors who sold IBM stock on Monday may be pricing in a future that is still technically unproven, much as some early assessments of robotaxi performance have struggled to capture the messy reality of large-scale deployment.
Could IBM Turn the Threat Into a Service Line?
There is an alternative reading of the situation that the market largely ignored on Monday. If AI-driven COBOL modernization becomes viable, the companies best positioned to deploy it at scale may be the same firms that already have deep relationships with COBOL-dependent clients. IBM’s consulting arm has decades of institutional knowledge about specific client codebases, compliance requirements, and system architectures. An AI tool, no matter how capable, still needs to be integrated, tested, and validated within those environments. IBM could, in theory, adopt or build competing AI migration tools and sell them as part of its existing services contracts, capturing the efficiency gains rather than losing revenue to them.
That scenario depends on speed. IBM’s competitive footing can erode quickly if clients begin experimenting with Anthropic’s tool independently, bypassing IBM’s consulting teams altogether. The company’s 2024 SEC filing acknowledges that emerging technologies pose risks to its software and infrastructure segments, but it does not outline a specific contingency for AI-automated code migration. That silence may have contributed to investor anxiety. Without a clear public response from IBM executives explaining how they will incorporate or counter these tools, boards may feel pressure to explore alternatives, much as policymakers are reassessing long-entrenched frameworks like traditional zoning rules in light of new economic and technological pressures.
What the IBM Shock Says About Legacy Systems and AI
IBM’s plunge is about more than one company’s earnings multiple; it is a signal of how markets may treat any business built on maintaining hard-to-replace legacy infrastructure. Investors have watched AI systems write code, summarize legal documents, and generate marketing copy, but the COBOL announcement targets a different layer of the stack: the mission-critical back-end systems that governments and financial institutions rely on every day. If automation can credibly move into that territory, the valuation of firms whose margins depend on scarcity of specialized skills will need to be revisited. The reaction to IBM suggests markets are ready to make those adjustments abruptly rather than gradually.
At the same time, history suggests that technological shifts often create as many service opportunities as they destroy. The Shaker religious communities profiled in recent cultural reporting offer a reminder that systems assumed to be fading can experience renewed relevance when outside interest and new tools converge. In the enterprise context, AI-enabled modernization could spur a wave of long-delayed upgrades, audits, and compliance projects that still require human oversight, governance, and integration work. Firms like IBM could find themselves orchestrating these transitions rather than merely defending old revenue streams, provided they move quickly enough to claim that role.
Regulation, Risk, and the Next Phase of AI Modernization
Another dimension largely absent from the immediate trading reaction is regulation. Banking and government systems are subject to stringent oversight, and any large-scale automated code migration will face scrutiny from auditors and regulators. Agencies that are already grappling with questions about automation in areas as diverse as financial markets and immigration enforcement are unlikely to accept black-box transformations of critical infrastructure without detailed validation. The political debate over issues like detention policy shows how quickly technology questions can become entangled with broader concerns about accountability and control.
For IBM, Anthropic, and their clients, that means the real contest may play out less in flashy demos and more in the slow work of building trust with regulators, insurers, and risk officers. Legacy systems are deeply intertwined with compliance regimes, from audit trails to reporting standards, and any AI modernization tool will have to prove not only that it works, but that it can be governed. As debates over how to modernize long-standing structures (from software stacks to urban planning rules) intensify, IBM’s stock shock may be remembered less as an overreaction and more as an early indicator of how abruptly markets will reprice incumbents when AI appears to challenge the foundations of their most protected businesses.
More From The Daily Overview
*This article was researched with the help of AI, with human editors creating the final content.
Grant Mercer covers market dynamics, business trends, and the economic forces driving growth across industries. His analysis connects macro movements with real-world implications for investors, entrepreneurs, and professionals. Through his work at The Daily Overview, Grant helps readers understand how markets function and where opportunities may emerge.
