On Monday, Anthropic published a blog post about using Claude Code to modernize COBOL. IBM’s stock dropped 13% — its worst single-day loss since October 2000. Bloomberg reported the decline put IBM on track for a 27% drop in February, its worst monthly slide since at least 1968.
Accenture and Cognizant also fell. The selling continued a pattern from Friday, when Anthropic’s Claude Code Security announcement sent cybersecurity stocks tumbling. In less than a week, Anthropic blog posts have triggered two separate market selloffs across entirely different sectors.
But here’s the thing. What Anthropic actually announced is a playbook for using AI to help analyze and refactor COBOL code. That’s a real capability. It’s also something IBM, OpenAI, AWS, Microsoft, and every major systems integrator has been saying — and shipping — for years.
What Anthropic Actually Said
Anthropic’s blog post makes a straightforward case. COBOL handles 95% of U.S. ATM transactions. Hundreds of billions of lines run in production. The developers who built these systems have retired. Universities don’t teach COBOL. Modernization has stalled because understanding the code costs more than rewriting it.
Claude Code, Anthropic says, can automate the exploration and analysis phases — mapping dependencies, documenting workflows, and identifying risks. The company published a Code Modernization Playbook alongside the post.
None of this is wrong. But the framing — that this turns years of modernization into quarters — left out important context.
This Isn’t New
IBM launched watsonx Code Assistant for Z in August 2023. It uses a 20-billion-parameter model trained on COBOL-Java pairs to help developers selectively refactor COBOL into Java — including application discovery, dependency mapping, automated refactoring, and validation. CEO Arvind Krishna said in July 2025 that the tool “has got very wide adoption.” And just last month, IBM reported its highest mainframe revenue in 20 years.
IBM isn’t alone. AWS has mainframe migration programs. Microsoft offers COBOL modernization tools. Kyndryl runs large-scale migration projects. Infosys chairman Nandan Nilekani said just last week that AI has made rewriting legacy apps affordable.
The idea that AI can help with COBOL modernization is well-established. What’s new is that a consumer-facing AI company said it publicly during a market already rattled by AI disruption fears.
What IBM Got Right in Its Response
IBM SVP Rob Thomas published a response on Monday. His core argument: translating COBOL isn’t the same as modernizing enterprise systems. The language isn’t the source of mainframe value. The platform is.
Thomas noted that 40% of COBOL doesn’t even run on mainframes. And COBOL on IBM Z represents decades of tight coupling between software and hardware — processor-level acceleration, I/O optimization, and performance tuning that doesn’t transfer when you move the code. His analogy: it’s like iOS and iPhone. The performance comes from hardware-software integration.
That’s a bit of spin — Java runs on mainframes, and IBM’s own tools convert COBOL to Java on Z, but the point holds. The hard part isn’t reading or translating the code. It’s data architecture redesign, transaction processing integrity, runtime replacement, and proving the new system does exactly what the old one did.
“Anthropic’s post on COBOL modernization certainly grabbed the market’s attention,” said Mitch Ashley, VP and practice lead for software lifecycle engineering at The Futurum Group. “But that conversation is only one part of what it takes to modernize applications. It also requires business scoping, behavioral equivalence validation, data migration strategy, and organizational change. The best team for serious COBOL modernization programs may include both vendors. That’s a different conclusion than the stock market reached on Monday.”
Evercore ISI analyst Amit Daryanani made a similar point, noting that IBM has already provided customers with several modernization options and that clients who had the option to migrate from the mainframe have been sticking with the platform.
The Real Story
Two things can be true at once. AI tools, including Claude Code, are genuinely useful for understanding legacy COBOL codebases. And a blog post about that capability doesn’t threaten IBM’s mainframe business the way a 13% stock drop implies.
The market is in “sell first, ask questions later” mode on AI disruption. A major software ETF is down 27% this year, on pace for its worst quarter since 2008. Every time Anthropic, OpenAI, or Google ships a new AI capability, investors recalibrate their expectations for whichever company is closest to the blast radius.
For DevOps teams working on actual mainframe modernization, the signal-to-noise ratio matters. AI-assisted code analysis is useful. But modernizing a COBOL system that processes billions of daily transactions requires business scoping, data migration, behavioral equivalence validation, regulatory compliance, and organizational change. No playbook eliminates that complexity.
The best approach probably isn’t Anthropic or IBM. It’s both — along with the systems integrators, testing tools, and domain experts who understand why these systems were built the way they were.
Wall Street overreacted on Monday. But the underlying question — how AI changes the economics of legacy modernization — is worth tracking. Just not at the pace the stock market implied.
