The AI free lunch is over: Model economics for platform engineers

IDC predicts organizations will underestimate AI infrastructure costs by 30% through 2027. Developers are selecting the most powerful models for trivial tasks, feeding entire codebases as context, and agentic tools have an insatiable appetite for tokens.

Sound familiar? Platform teams have seen this before. In this talk, Bryan shows how the FinOps controls that tamed public cloud spend can stop organizations from repeating history with AI.

Key takeaways:
• Why AI costs are unpredictable, non-linear, and invisible to end users.
• Practical platform patterns: model routing, context caching, cost attribution, and spending guardrails.
• How to build cost-aware choices into a platform without slowing developers down.