- Anthropic afirma que Claude Code puede acelerar la modernización de COBOL automatizando tareas de exploración y análisis que antes requerían grandes equipos de consultores.
- Las acciones de IBM sufrieron una fuerte caída intradía ante el temor de que la nueva herramienta de IA ponga en cuestión parte de su negocio ligado a mainframes COBOL.
- El anuncio se suma a una oleada de presión sobre el sector software, con un ETF de software acumulando un retroceso significativo en el año.
- Inversores temen que la IA para escribir código reduzca la demanda de soluciones tradicionales, afectando crecimiento, márgenes y capacidad de fijar precios.
In recent trading sessions, Anthropic’s new COBOL-focused capabilities in its Claude Code tool have become an unexpected catalyst for volatility across parts of the technology market. What might seem like a niche update aimed at modernizing an old programming language is, in practice, stirring broader concerns about how artificial intelligence could reshape long-standing business models, especially those tied to legacy infrastructure.
Investors, analysts and software teams are now trying to understand how an AI system that promises to streamline COBOL modernization work could affect companies that rely heavily on mainframe-based revenue. The immediate market reaction has been particularly visible in IBM shares and in a wide swath of software and cybersecurity stocks, highlighting just how sensitive the sector has become to advances in code-generating AI.
Anthropic’s Claude Code targets COBOL modernization
According to Anthropic, Claude Code is designed to simplify the notoriously complex process of upgrading COBOL systems, which still underpin critical operations in industries like finance and government. COBOL, despite its age, remains a backbone language on many enterprise mainframes, and updating those systems has historically been a slow and expensive undertaking.
The company explained in a blog post that modernizing a COBOL environment used to mean deploying large teams of consultants for years, painstakingly mapping workflows, dependencies and data flows. Much of the effort went not into writing new code, but into understanding what the existing applications actually did and how they interacted with surrounding systems.
Anthropic now claims that Claude Code can automate big chunks of the exploration and analysis phases that consume most of the time and budget in a typical COBOL modernization project. By analyzing source code, documentation and system behavior, the tool aims to generate structured insights that would previously have required manual investigation.
This approach, if it delivers in real-world environments, could mean that organizations spend fewer resources on initial discovery and mapping, potentially shortening modernization timelines and reducing reliance on large consulting engagements focused purely on legacy analysis.
At the same time, the move underscores how AI is increasingly being positioned as a way to bridge the gap between decades-old codebases and modern architectures. Instead of rewriting everything by hand or relying solely on specialized legacy experts, companies may lean on tools like Claude Code to accelerate and de-risk the transition.
Why IBM felt the impact so sharply
One of the most immediate ripple effects of Anthropic’s announcement appeared in the share price of International Business Machines Corp. IBM’s stock tumbled sharply, with an intraday drop exceeding double-digit percentages, marking its steepest move of that magnitude since the early months of the pandemic in 2020.
Over the month, the decline deepened, leaving IBM on track for its most severe percentage loss in a single month in decades, based on historical data compiled by Bloomberg going back to the late 1960s. That kind of move is unusual for a company often perceived as a stable, legacy technology player.
The market’s reaction is closely tied to IBM’s enduring dependence on its mainframe division. A notable slice of the company’s revenue and client relationships still revolves around these large servers, which are typically owned and operated by customers running mission-critical applications in COBOL.
For banks, insurers and government agencies, IBM mainframes are valued for their reliability, throughput and security. The systems support high-volume transaction processing, regulatory workloads and other operations where downtime or errors would be unacceptable. Because of that, modernization projects have tended to be cautious, slow and heavily supported by specialized IBM services and partners.
Anthropic’s claim that Claude Code can take on the time-consuming groundwork of COBOL modernization sparked concerns that some of that high-margin, services-heavy activity could face new competitive pressures. Even if mainframes remain in place for years, investors are asking whether the long-term revenue profile tied to legacy maintenance might look different in a world where AI plays a bigger role.
Legacy systems, AI and a shifting consulting landscape
Beyond the immediate stock move, the prospect of automating deep analysis of COBOL systems touches a broader ecosystem of consulting firms, systems integrators and niche vendors that have built businesses around legacy modernization.
Traditionally, modernization projects have followed a familiar pattern: assemble a team of experts, audit the existing code and data, document current workflows, then design a step-by-step migration plan. This kind of work can stretch over multi-year engagements, with significant consulting fees attached.
By positioning Claude Code as a tool that can rapidly parse large COBOL codebases and surface insights, Anthropic is implicitly suggesting that some of the manual labor in these projects might be reduced or reframed. Instead of spending months on basic discovery, teams could theoretically move more quickly into architecture decisions and implementation.
That does not necessarily eliminate the need for human expertise. Complex regulatory, security and organizational requirements still require seasoned professionals who understand both the legacy environment and the target architecture. However, the balance between manual code analysis and higher-level design work could shift if AI tools prove reliable at scale.
For organizations that have delayed modernization due to cost or complexity, the availability of AI-assisted COBOL tools could make long-postponed projects more feasible. At the same time, firms that derive a large share of revenue from traditional discovery and documentation work may need to adjust their value propositions to emphasize strategy, governance and integration.
Security features in Claude and the reaction in cybersecurity stocks
Shortly before the market turmoil around IBM and COBOL, Anthropic rolled out a new safety feature in its Claude AI model, aimed at reinforcing secure and responsible behavior when generating content or code. That update, while focused on risk mitigation, also had an unexpected impact on another corner of the market: cybersecurity stocks.
Following the announcement, there was a broad selloff across several cybersecurity names, as traders digested the idea that more capable, safety-conscious AI tools might change how organizations approach certain security functions. The move added to a year in which software valuations have already been under pressure.
Market data show that a major software-focused ETF has dropped markedly this year, putting it on pace for its sharpest quarterly decline since the global financial crisis in 2008. That context helps explain why any notable AI announcement—whether related to code generation, security or productivity—tends to trigger an outsized response.
Investors appear to be weighing whether increasingly capable AI assistants could absorb some tasks currently performed by dedicated security tools or services. While there is no consensus on how quickly that might happen, the recurring pattern of selloffs suggests a high level of sensitivity to the possibility.
At the same time, some market participants argue that AI-enhanced defenses could just as easily complement existing cybersecurity stacks, driving new demand for tools that can orchestrate, monitor and govern AI-generated actions. For now, though, the dominant reaction in share prices has skewed toward caution.
Code-writing AI and fears of disruption in traditional software
Anthropic’s work on COBOL is part of a broader trend in which AI models from Anthropic, OpenAI, Alphabet and others are increasingly able to generate software code. These tools can write, refactor and explain code in multiple languages, lowering the barrier to building applications and automations.
That capability—often referred to as the ability to “code” using AI—has raised concerns that users may rely more on AI assistants instead of purchasing conventional software products. If organizations can quickly prototype internal tools or tailor-made workflows using generative AI, they might feel less need for off-the-shelf applications in certain categories.
From an investor standpoint, this leads to questions about future growth, margins and pricing power for established software vendors. If AI lowers development costs and expands the number of people who can build solutions, competitive dynamics could shift, potentially compressing the premium that some companies have been able to command.
These worries have contributed to renewed volatility in software equities, as each major AI release is scrutinized for its potential to erode traditional revenue streams. The reaction to Anthropic’s COBOL-focused capabilities fits squarely into that pattern, even though the immediate use case targets a relatively specialized domain.
At the same time, there is an opposing view: AI-generated code might spur new demand for infrastructure, security, integration and governance tools, as organizations look to manage a growing volume of software artifacts produced by both humans and machines. How that balance plays out will be central to the sector’s longer-term trajectory.
Ultimately, Anthropic’s COBOL tool has become a focal point in a broader debate about how generative AI will interact with legacy technology, consulting models and the software industry at large. The sharp moves in IBM and cybersecurity stocks underscore how much is at stake, but the long-run impact will depend on how reliably tools like Claude Code perform in complex, real-world modernizations and how quickly enterprises adapt their strategies in response.
