Most conversations about AI focus on models, capabilities, and use cases. This episode goes into the financial planning and plumbing underneath the entire AI economy. Anastassia sits down with Carmen Li — a former Bloomberg and Citi executive turned founder — to unpack GPU compute cost volatility. Every AI application, every model inference, every startup scaling its product runs on compute — and yet there is almost no financial infrastructure to benchmark, price fairly, or hedge against the wild swings in GPU costs.

Carmen built the world's first GPU compute index, published it on the Bloomberg Terminal within months of founding Silicon Data, and is now building Compute Exchange — a marketplace where compute can be traded as transparently as oil, electricity, or any other commodity. Together, Carmen and Anastassia explore why compute is not just a cost but a strategic resource, why AI companies are flying blind without proper risk management tools, how geopolitical tensions are bifurcating the global chip market, what the rise of open-source models means for European and mid-sized businesses, and how Carmen raised $5.6 million without a pitch deck. A masterclass in the economics behind the AI revolution.

Chapters:

00:04 Introduction — GPU Compute: The Wild West of AI Finance

02:18 Carmen Li and The Trillion-Dollar Blind Spot

02:50 Why Compute Needs the Same Infrastructure as Oil and Energy

04:33 AI Runs on Compute, and Compute Costs a Fortune

05:39 Carmen's Journey — From Trading Floors to Silicon Valley

08:01 The Problem: GPU Cost Volatility Is Breaking AI Startups

11:07 Vision for the Future — Compute CapEx Over 10–15 Years

11:50 The GPU Compute Index on Bloomberg and What's Launching Next

13:37 The LLM Expenditure Index — Token Costs Are Actually Rising 38%

16:16 Compute as Strategic Resource — Not Just a Cost Line

18:00 Semiconductor Industry and their insights