IBM's sharp selloff underscores how artificial intelligence (AI) is beginning to challenge long-standing technology moats in enterprise computing, raising broader questions about who controls modernization budgets and the future architecture of corporate IT systems.
Shares of IBM suffered their steepest single-day drop in more than two decades after AI startup Anthropic said its Claude Code tool could help automate the modernization of COBOL systems that run heavily on IBM mainframes. The stock fell about 13% in one session, capping a monthly slide of roughly 27%, as investors reacted to the possibility that AI could erode one of IBM's traditional strongholds, according to Market Watch and The Investor's Business Daily.
Anthropic said Claude Code can automate the labor-intensive exploration and analysis phases of COBOL modernization — work that historically required large teams of consultants mapping complex workflows across decades-old systems. COBOL, first released in 1960, remains deeply embedded in financial services, airlines, and government systems, with an estimated 95% of US ATM transactions relying on it. IBM manufactures the majority of mainframes running these workloads, and modernization services have long been intertwined with its infrastructure and consulting revenue.
The market response reflected concern that AI-native tools may compress the time and cost required to migrate or refactor legacy code, potentially weakening a business model built around specialized expertise and long project cycles. Software stocks have been under pressure this year amid fears that generative AI could undermine traditional licensing, maintenance, and consulting revenue streams.
Yet analysts cautioned that the immediate selloff may overstate the threat. IBM has introduced its own AI-powered tools through its WatsonX platform to assist in refactoring COBOL into modern languages such as Java. Executives have argued that customers already had options to migrate off mainframes, but often remain due to reliability, security, and regulatory requirements. For mission-critical environments, especially in banking and government, modernization tends to be incremental rather than wholesale.
From mainframes to AI agents
Beyond the stock volatility, the episode highlights a deeper technology transition: the shift from platform-centric dominance to model-centric leverage. For decades, control over hardware and proprietary software ecosystems defined enterprise power. IBM's mainframes exemplified that model — vertically integrated systems optimized for large-scale transaction processing.
AI changes the equation by abstracting complexity at the software layer. Tools like Claude Code promise to interpret and refactor millions of lines of legacy code with far less manual effort. If successful, that could reduce switching costs and weaken infrastructure lock-in. The perceived value shifts from owning the underlying platform to commanding the most capable AI models and orchestration layers.
Similar paradigm anxieties have surfaced before. When OpenAI launched ChatGPT in late 2022, shares of Alphabet Inc. fell sharply amid fears that generative AI could disrupt the search advertising model. Although Google responded by accelerating its own AI rollout, the episode marked a moment when investors questioned whether model innovation could reshape a long-entrenched business.
Earlier, the rise of cloud-native software companies such as Salesforce pressured legacy vendors, including Oracle Corporation and SAP, as subscription-based delivery threatened perpetual licensing models. Over time, incumbents adapted, but not before markets repriced their growth prospects to account for a structural shift.
In IBM's case, the dynamic is more complex because Anthropic is both partner and potential disruptor. The two companies have collaborated to integrate Claude into IBM's enterprise offerings, positioning IBM as a secure, governance-focused bridge between cutting-edge models and risk-averse industries. At the same time, Anthropic's aggressive push into legacy modernization strikes at a revenue stream that has historically reinforced IBM's customer relationships.
Ultimately, the question is less about whether COBOL disappears and more about who controls the modernization layer. If AI agents become the primary interface for rewriting, auditing and securing enterprise code, value may migrate toward model developers and orchestration platforms rather than hardware vendors. IBM's challenge — and opportunity — lies in embedding itself within that new stack rather than defending the old one.
Analysts argue selloff overstates modernization risks
Analysts at Evercore ISI said fears over potential mainframe migration are understandable but overstated, noting that IBM has already equipped customers with multiple modernization pathways and that many clients continue to stick with the platform because of its reliability and other structural advantages. The firm described the recent stock drop as "unwarranted."
Jefferies echoed that view, arguing that concerns ignore the fact IBM is actively disrupting its own legacy model through watsonx Code Assistant for Z, which embeds generative AI directly into mainframes to refactor COBOL into Java and modernize applications with full system context. Jefferies added that the resilience of the mainframe business is "far greater than many assume."
Article edited by Jack Wu
