Artificial intelligence is becoming cheaper to operate as model efficiency improves and inference costs decline, but those savings may disguise a much larger expense created by widespread adoption: cybersecurity. As businesses deploy autonomous AI agents across corporate networks, databases, applications and sensitive information, the number of potential attack surfaces multiplies dramatically. Estimates cited in the analysis suggest agentic AI could increase token consumption roughly 24-fold by 2030, while security spending associated with the expanding AI ecosystem could ultimately add hundreds of billions of dollars. Independent research supports the broader thesis: cybersecurity budgets are already rising rapidly, organizations remain concerned that defenses are failing to keep pace with AI-enabled threats, and specialized AI-security spending is accelerating. The result is an economic contradiction Wall Street may be underestimating: intelligence can become cheaper at the same time the infrastructure required to trust, govern and protect it becomes considerably more expensive.
Key Takeaways
- Falling AI inference costs are likely to stimulate substantially greater usage rather than simply reduce corporate technology expenses, particularly as autonomous agents become embedded throughout business operations.
- AI dramatically expands the cybersecurity perimeter by introducing models, agents, prompts, training data, AI-generated code and non-human identities that companies must secure alongside traditional networks, applications and data.
- Cybersecurity could become one of the largest secondary beneficiaries of the AI investment cycle, as businesses discover that cheaper artificial intelligence does not eliminate the increasingly expensive requirement to secure it.
In-Depth
Artificial intelligence may be getting cheaper to run, but that does not mean the total cost of deploying it is falling. Lower inference prices encourage companies to automate more tasks, deploy more agents and process vastly more data. One estimate cited in the underlying analysis suggests agentic AI could drive roughly 24 times more token consumption by 2030, with enterprise agents producing an even steeper increase.
That expansion creates a less celebrated expense: security. As AI agents gain permission to access corporate systems, customer information, proprietary data and operational tools, every new connection creates another potential point of attack. Cybersecurity therefore becomes less of an optional technology expense and more of a prerequisite for scaling AI safely.
Recent industry research reinforces the concern. A survey of cybersecurity leaders found 83 percent are increasing cyber spending, while nearly nine in ten reported attacks during the previous year. Separate research found AI-related cybersecurity already accounts for more than 11 percent of cybersecurity budgets at surveyed enterprises, with most respondents increasing investments in AI-specific defenses.
The investment implication is straightforward. Falling model and inference costs may improve the economics of AI itself, but investors who focus only on compute efficiency risk overlooking the cost of protecting the resulting infrastructure. The AI boom is not simply a race for cheaper intelligence; it is also creating an enormous new security perimeter. That dynamic could redirect hundreds of billions of dollars toward cybersecurity over the coming years, making security providers important beneficiaries of widespread AI adoption.
Sources
- https://www.bcg.com/publications/2026/cybersecurity-spending-ai-threat-trends
- https://www.gartner.com/en/newsroom/press-releases/2026-08-26-gartner-forecasts-the-market-for-securing-ai-will-reach-almost-5-billion-in-2027
- https://ir.isg-one.com/news-market-information/press-releases/news-details/2026/Enterprises-Boost-AI-Cybersecurity-Spending-But-Fear-Investments-Lag-Emerging-Threats-ISG-Study/default.aspx
- https://www.gartner.com/en/newsroom/press-releases/2026-09-16-gartner-forecasts-worldwide-ai-spending-to-grow-49-point-5-percent-in-2026