Moving past the cost-cutting narrative to build sustainable AI infrastructure.

I'm documenting my work at AssemblyAI to help shape how the AI industry defines unit economics, a perspective that's almost entirely absent from the conversation today.

"For AI infrastructure providers, relying on the monthly invoices provided by cloud providers is structurally insufficient to support the decisions Engineering, Sales, and Finance need to make. Accurate cost-of-goods-sold (COGS) and unit economics for an AI infrastructure company requires a methodology that joins multiple incomplete billing data sources to fill in crucial gaps in reporting, allocates compute cost into causally distinct layers, and buckets costs into capacity classes highlighting areas of opportunities for cost efficiencies. This process involved incorporating documented methodology from credible sources - such as AWS - in addition to creating proprietary methodology designed for AssemblyAI. The purpose of this post is to share how we went about this as part of an effort to help shape how the AI industry defines unit economics."

Building the Foundation for Unit Economics at an AI Infrastructure Company →