The Sticker Shock of AI Infrastructure Costs
13m
Governance Meets the AI Era
AI infrastructure costs are about to climb, and many organizations don't see it coming. AI agents will consume far more resources than today's workloads. Most teams only notice when the first bill arrives.
CloudBolt CEO Alexander Thomson examines that challenge with Mike Vizard. Thomson recently took the helm at CloudBolt after serving as COO at PlexTrac. He says the core mission stays the same: help enterprises and service providers govern, manage and automate their resources. What's changing is the scope. After private and public cloud, the next wave is agentic workloads and MCP agents running on Kubernetes.
From Tickets to Automation
Today, infrastructure teams still want a human in the loop. They act on recommendations but hesitate to hand agents full control. Thomson saw the same pattern in cybersecurity.
He expects that to change. Tickets in ServiceNow or Jira will remain, but mostly for audit. In a Kubernetes world, a two-week lag on rightsizing is too slow.
Policies alone won't stop shadow AI, he adds. Teams need room to self-serve, while IT keeps visibility and control in the background.
Bringing Cost Into the Workflow
Engineers make better choices when they can see what something will cost. Thomson says most people don't want another dashboard. They want data where they already work. MCP connectors can surface forecasted costs as resources are deployed.
Trust in that data matters too. When teams doubt the numbers, friction follows.
A Dirty Secret in the Making
Thomson calls rising AI infrastructure costs a dirty secret that hasn't fully surfaced. GPU workloads cost more, and energy adds to the bill. Frontier model companies are also offering credits to win market share. As they chase margins, prices will likely rise.
His advice to IT leaders is to embrace adoption rather than block it. Build guardrails instead of walls. Plan for efficiency up front to avoid sticker shock later.
Thomson wants CloudBolt to be the foundation for governance in the AI age.