One blog post and a research preview. That’s all it took for Anthropic to wipe over $45 billion from the combined market caps of IBM, CrowdStrike, and other enterprise tech giants on February 23, 2026.
IBM suffered its worst single-day loss since October 2000, plunging 13.2% after Anthropic published a Code Modernization Playbook demonstrating how Claude Code can tackle COBOL modernization - a multi-billion dollar consulting business that IBM has dominated for decades.
Hours later, cybersecurity stocks followed IBM off a cliff when Anthropic debuted Claude Code Security in a limited research preview. CrowdStrike dropped 11%, Zscaler fell 10%, and the Global X Cybersecurity ETF tumbled 4%.
Welcome to the Anthropic Effect.
The COBOL Bomb
Hundreds of billions of lines of COBOL still run in production today. Banks, airlines, insurance companies, and government agencies depend on code written decades ago, maintained by an aging workforce that’s literally dying off.
IBM built an empire on this technical debt. Its legacy modernization consulting business charges premium rates to translate ancient COBOL systems into modern code - projects that typically take years and cost millions.
Then Anthropic dropped this line in their blog post: “With AI, teams can modernize their COBOL code base in quarters instead of years.”
Claude Code can map dependencies across thousands of lines of COBOL, document hidden workflows, identify migration risks, and provide analysis that would take human consultants months to surface. The tool builds scaffolding to run old and new code simultaneously during transitions, letting organizations migrate component by component.
“AI code translation directly competes with IBM’s legacy modernization consulting,” noted one analyst. The market agreed, wiping $31 billion from IBM’s market cap in hours.
With the decline, IBM shares have fallen 27% in February alone - on track for its biggest one-month slide since at least 1968.
The Security Shockwave
The COBOL announcement was the morning hit. The cybersecurity bloodbath came in the afternoon.
Anthropic unveiled Claude Code Security as a limited research preview, describing it as a tool that can scan software code for vulnerabilities and suggest patches. Nothing revolutionary on paper - but the market saw an existential threat.
CrowdStrike tumbled 11.3%, marking its most significant single-day drop in years. The panic spread across the sector: Zscaler fell 10%, Fortinet and Okta dropped more than 5%, Datadog slid alongside them.
The fear isn’t about what Claude Code Security does today. It’s about what it implies for tomorrow.
AI-native startups have begun unveiling autonomous “remediation” tools that don’t just detect vulnerabilities - they fix them automatically. If AI can patch security flaws before they’re exploited, what happens to the high-margin “detection and response” services that companies like CrowdStrike sell?
The CEO Response
CrowdStrike CEO George Kurtz rushed to defend his company’s moat: “An AI capability that scans code does not replace the Falcon platform. Security requires an independent, battle-tested platform built to stop breaches.”
Not everyone agreed with the market’s doom-and-gloom verdict. Wedbush analyst Dan Ives characterized the selloff as a “ghost trade,” calling CrowdStrike the “gold standard of cybersecurity” and arguing its Falcon platform is well-positioned against AI-driven threats.
But the market had already voted with $45 billion.
What This Means
This isn’t the first time AI announcements have moved markets. DeepSeek’s efficiency breakthroughs crashed Nvidia 17% in January 2025. OpenAI’s Sora preview sent video editing software stocks tumbling.
But the Anthropic Effect feels different. These weren’t demos of future capabilities - they were practical tools targeting specific, identifiable revenue streams.
IBM’s COBOL consulting. CrowdStrike’s security services. Both represent multi-billion dollar businesses that depend on human expertise being expensive and scarce.
Claude Code attacks both assumptions simultaneously.
For enterprises, this creates genuine opportunity. COBOL modernization has been so expensive that many organizations simply accepted the risk of running decades-old code. If AI can cut those costs by 80% while reducing timeline from years to quarters, we might finally see the mass migration that should have happened a decade ago.
For workers in these industries, the math is less encouraging. IBM employs thousands of consultants who specialize in legacy modernization. CrowdStrike and its competitors employ security analysts whose job is exactly what Claude Code Security promises to automate.
The Privacy Angle
Lost in the market chaos: what happens when AI tools start scanning enterprise codebases?
Claude Code’s COBOL analysis requires access to source code that runs critical financial infrastructure. The security tool needs to see vulnerabilities before they’re patched. Both represent significant trust decisions for enterprises.
Anthropic’s enterprise products operate under strict data handling agreements. But the broader industry question remains: as AI tools need more access to do their jobs, how do organizations balance capability against exposure?
The Microsoft Copilot incident from earlier this month - where a bug let the AI summarize confidential emails despite DLP labels - shows these aren’t hypothetical concerns.
What You Can Do
If you work in enterprise consulting or cybersecurity:
- The disruption is real but not immediate. Claude Code requires human oversight; it’s not replacing entire departments tomorrow
- Shift your expertise toward AI orchestration. Knowing COBOL becomes more valuable, not less, when you can direct AI tools effectively
- Watch the enterprise adoption curve. The research preview to general availability gap is where competitive advantage gets built
If you’re in an organization running legacy systems:
- Get quotes from traditional consultants now. They’re about to get a lot cheaper, or disappear
- Evaluate Claude Code’s actual capabilities against your specific codebase. Marketing claims and real-world performance often diverge
- Consider the security implications of AI-assisted modernization. Faster isn’t always better if it introduces new vulnerabilities
If you’re an investor:
- The Anthropic Effect isn’t going away. Every new capability announcement will be parsed for its disruption potential
- Traditional moats are being tested. “Expertise is expensive” used to be a defensible business. Now it’s a vulnerability
- Watch for overreaction. Dan Ives might be right about CrowdStrike being oversold. AI announcements create volatility opportunities in both directions
February 23, 2026, might be remembered as the day the market finally priced in AI’s impact on enterprise software. Or it might be remembered as an overreaction that created buying opportunities.
Either way, the Anthropic Effect is now a market reality. Every AI lab announcement will be measured against the question: whose lunch does this eat?