LIVE WORKSHOP
Your Platform Team Isn't Your Biggest Spender
Unit economics and cost to serve when shared infrastructure is most of the bill
Look at who carries the biggest line in your showback. For most organizations it's the platform team, because it runs the shared clusters, the data platform, and the model gateway. Every dollar is tagged, every rule was agreed, and the total matches the invoice. It's still wrong, because the platform team owns that infrastructure and other teams consume it.
Tags measure ownership. For fifteen years that was close enough, because whoever provisioned a resource was whoever used it. Shared infrastructure broke that assumption, and AI is turning shared infrastructure into most of the bill: one team runs the endpoint, while a dozen teams and thousands of customers call it.
In this live workshop, Amit Kinha, Field CTO at DoiT and FinOps Foundation Governing Board member, takes a real cloud bill and shows you how to re-cut it by consumption instead of ownership. The total stays the same and the owners change. He'll walk through where rule-based allocation stops working, how fast its error grows as shared cost grows, and what it takes to produce unit economics and cost-to-serve numbers Finance will defend.
What you'll learn
- Why tag-based showback can match the invoice to the dollar and still assign cost to the wrong teams
- Where rule-based allocation stops working, and how fast its error grows as shared infrastructure and AI become most of the bill
- How to re-cut a real bill by consumption instead of ownership, keeping the total the same while the owners change
- What it takes to produce unit economics and cost-to-serve numbers that Finance will defend
Meet the experts
Amit Kinha
Field CTO, DoiT
Amit Kinha is Field CTO at DoiT, where he works with large enterprises on the hard parts of running cloud and AI at scale. He is a Governing Board Member of the FinOps Foundation and a founding member of the Tokenomics Foundation. Previously, Amit was Global Head of Cloud FinOps at Citi, managing roughly $150M a year in spend, and spent seven years in backend engineering at Goldman Sachs on Marcus and Apple Card. He co-hosts the Applied FinOps podcast and speaks regularly at AWS re:Invent, Google Cloud Next, and FinOps X.
More resources
Get ahead of the session with the reading that informs it.
5 Reasons AI Costs Don't Add Up (and How to Fix It)
Why every chargeback program has an "unallocated" line, why AI is making it grow, and how to attribute shared model APIs and GPU clusters without tags or code changes.
Get the ebook
Cost Per Customer: A FinOps Field Guide to Multi-Tenant Cloud Attribution
Working SQL for CUR 2.0, BigQuery billing exports, and FOCUS, plus a clear line on where native tooling stops and shared infrastructure needs measured consumption.
Get the guide
Cost Per Task, Not Cost Per Token: Unit Economics of Frontier LLM Workloads
Why per-token pricing hides the real cost of LLM work, and how to measure the unit Finance and Engineering can both agree on.
Read the blog
Your Agent-as-a-Service Bill Arrived. Now Explain the Unit Economics
A hands-on walkthrough of per-tenant cost attribution for Bedrock AgentCore, including a 280x spread in cost-to-serve between tasks on the same bill.
Read the blog