Column|Cybersecurity: AI Expands Demand for Remediation and Continuous Protection

Security could become a substantial source of AI model revenue, with token spending potentially reaching $50–70b as remediation cycles accelerate. The market would remain smaller than the coding opportunity, but still substantial.

AI labs are likely to capture most of this market, but if 20%–30% of the value accrues to security vendors, their opportunity from model revenue alone would be $10–20b. That compares with combined 2026 revenue of approximately $25b for the three security platform vendors, Cloudflare (NET), Palo Alto Networks (PANW), and CrowdStrike (CRWD).

In this note, we review and respond to recent comments from executives at AI labs and cybersecurity companies on security threats, including the PHP vulnerabilities. We draw certain implications from this discussion, including support for the $50–70b model revenue opportunity sized above.

The incremental model revenue opportunity from security

The estimate starts with the June 2026 Anthropic Economic Index dataset: cybersecurity and threat detection accounted for 1.42% of sampled first-party API requests in April and 1.52% in May, excluding Claude Code.

As security tasks require advanced models with at least Mythos-level capabilities, we estimate 1.5% share of API usage corresponds to a 4%–6% share of revenue. With global AI model revenue of approximately $200b and an assumed API contribution of 60%, security-related revenue would be roughly $5–7b.