Скачать или смотреть The New Economics of AI. Managing Token Costs, Margins, and Model Efficiency at Scale

📍 Recorded live at AI INFRA SUMMIT 4, Convene San Francisco

As AI workloads surge, tokens are becoming the new unit of cost, value, and strategy. In this session, Carmen Li (Silicon Data, Compute Exchange) breaks down how teams can understand and manage token level economics across models, infrastructure, and user behavior.

Carmen shares lessons from running large scale AI workflows, where token spikes, reasoning overhead, and workflow design can make or break margins. She explains why visibility into token usage, logging, and model routing is now essential for any AI product team.

Highlights from the session:

Why token cost is replacing GPU hours as the primary pricing and forecasting metric
The hidden cost drivers inside AI workflows, from context windows to reasoning models
How to calculate cost per request and cost per user to understand true unit economics
Logging, observability, and cross team workflows for managing model spend
Practical strategies for optimization, routing, quantization, batching, and feature level forecasting

📣 Super early bird available — sign up for the next AI INFRA SUMMIT → https://luma.com/aiinfra5

#amd #podcast #machinelearning