FinOps for AI: Cost per Task, Not Cost per Token

Registration
Only registrations through the link, not on Meetup. Register here: https://gdg.community.dev/events/details/google-gdg-west-sweden-presents-finops-for-ai-cost-per-task-not-cost-per-token/

Your CFO asks if the AI spending is paying off. “I think so” won’t cut it.

Gartner expects 40% of agentic AI projects to be cancelled by 2027 because their return on investment can’t be defined. Most teams can see their token spend; far fewer can say which feature, team or customer caused it, or whether it paid off. This talk gives you a practical FinOps method for ML and GenAI on Google Cloud’s Gemini Enterprise Agent Platform (formerly Vertex AI).

We start with the classic ML pipeline and what drives the bill. Then we move to GenAI, covering serving modes, caching, model routing, and controls that cap runaway cost. Next comes attribution: labeling every call and job, where labels fall short, and how eBPF observes traffic at the kernel level with no code changes. We finish with budgets, anomaly detection, and the team habits that make cost reviews stick.

You will leave with a shift in metric, from cost per token to cost per completed task, and five things to do on Monday.

Schedule:

• 17:30 - Doors open
• 18:00 - Talk + Q&A with Google Developer Expert Joshua Fox
• 19:00–21:00 - Food, drinks & mingling

English. Light food and drinks provided.
Location: Devies offices (Vallgatan 14), Gothenburg.