Cybersecurity Sell-Off: AI Leak Triggers Sector-Wide Price Drop

The market's reaction was swift and severe. On Friday, cybersecurity stocks sold off sharply, with CrowdStrike dropping 7% and Palo Alto Networks declining 6%. Other major names followed, including Zscaler, Okta, SentinelOne, and Fortinet, all shedding 3% to 4.5%.

The trigger was clear: reports of a leaked Anthropic AI model. Details about the new tier, codenamed "Capybara," were found in a publicly-accessible data cache, first reported by Fortune. The leaked draft described the model as "far ahead of any other AI model in cyber capabilities".

This revelation sparked immediate fear. The draft warned the model could spark a "wave of models that can exploit vulnerabilities", directly threatening the core business of defending against cyberattacks. The sell-off was a direct price impact from that perceived existential risk.

The Sector's Underlying Flow Disconnect

The market's reaction reveals a stark disconnect between financial fundamentals and stock prices. While cybersecurity stocks sold off last week, the sector's underlying revenue flows tell a different story. Aggregate cybersecurity revenue grew 16.8% in 2025, and the median company's growth rate stabilized at 13.7%. Yet, the median stock declined 18% over the past year. This divergence is the core issue: strong, if slowing, business performance is not translating into positive equity returns.

CrowdStrike's recent results exemplify this disconnect. The company posted record quarterly results, with ARR surpassing $5 billion and free cash flow of $376 million. Despite beating expectations, shares fell after the report. This pattern is not isolated. The evidence shows that stock price changes have limited correlation with 2025 revenue and earnings outcomes. Instead, performance is increasingly tied to expectations for future growth and profitability metrics like the Rule of 40.

This repricing is driven by concentration and sentiment. The top three cybersecurity companies now account for 68% of total market capitalization, creating a pronounced "haves versus have-nots" dynamic. For the typical company, even solid fundamentals are overshadowed by broader software-sector valuation compression and a market focused on a handful of mega-caps. The sell-off is a symptom of that concentrated repricing, where flows are shifting away from the sector as a whole.

The IBM Shock: AI's Direct Threat to Legacy Tech

The threat from AI is no longer theoretical for established tech. IBM shares crashed as much as 13% on Monday after Anthropic announced its Claude Code tool can automate a core, labor-intensive part of upgrading COBOL systems. This is a direct hit to IBM's mainframe business, which supports applications written in the decades-old language that runs on its own hardware.

The mechanism is clear. Anthropic stated its tool can automate the "exploration and analysis phases that consume most of the effort in COBOL modernization." This directly threatens IBM's mainframe consulting services, a substantial portion of its operations. The stock's steepest single-day drop since 2020 and its 26% decline for February signal investors see this as a material growth risk.

This fits a broader "vibe coding" trend where AI generates code, potentially reducing reliance on expensive legacy software and consulting. For IBM, it's a double whammy: a new AI tool can accelerate the very modernization projects that have long been a source of its consulting revenue. The crash underscores that AI disruption is now targeting the profit centers of legacy tech giants.