AI’s Reckoning: $725B in Spend, Weak Utilization and New Security Risks
Hyperscalers may spend $725 billion on AI infrastructure in 2026, but the real story is not just how much money is being deployed. It is how much of that capacity is being used effectively, how quickly inference costs are becoming the next form of cloud sprawl and how agentic AI is expanding the security threat surface.
On this episode of Techstrong Gang, Jon Swartz, Mike Vizard and Stephen Foskett connect the dots across AI spending, infrastructure efficiency, knowledge graphs and Security Field Day 15 to unpack what these shifts mean for enterprise technology leaders.
The conversation starts with the growing tension between massive AI investment and weak real-world utilization. Even as hyperscalers and chipmakers pour capital into data centers and compute capacity, Kubernetes environments remain underutilized and GPU inference is emerging as a costly new operational blind spot. The issue is no longer just scale. It is discipline, efficiency and whether organizations can turn raw capacity into real business value.
The episode also explores the rise of knowledge graphs as a possible context layer for AI agents. As enterprises look for better ways to make agents more auditable, responsive and less dependent on brute-force token consumption, knowledge graphs are re-emerging as a serious architectural answer for grounding and governance.
Finally, the gang looks at the security implications surfacing at Security Field Day 15, where the focus is increasingly shifting toward the new threat models created by agentic systems. As AI agents gain access to more tools, workflows and decision loops, organizations are being forced to think more carefully about exposure, control and which security primitives will become essential in the next phase of enterprise AI.
Taken together, these stories point to the same core reality: the AI era is no longer just about spending more. It is about using infrastructure better, grounding intelligence more effectively and securing a much more autonomous technology stack.
