The $7B Startup Building AI Chips to Challenge Nvidia | Fractile Founder & CEO, Walter Goodwin
Source
youtube.com
Author
20VC with Harry Stebbings
Date
Why it matters
Covers why memory bandwidth, not compute, limits inferenceRunning a trained model to get answers — the phase where AI is actually used, as opposed to trained.Full definition → speed and what that implies for AI agentAn AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.Full definition → workloads and future tokenThe chunk of text a model reads and writes in — roughly three-quarters of a word — and the unit AI usage is billed in.Full definition → cost. Useful for engineers planning around inference economics and availability.
Terms in this piece · Glossary
inference — Running a trained model to get answers — the phase where AI is actually used, as opposed to trained.
AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
token — The chunk of text a model reads and writes in — roughly three-quarters of a word — and the unit AI usage is billed in.
videoMistral CEO: Why OpenAI & Anthropic Will Come for Your Business | Are Chinese Models Truly Safe?
videoThe Future of Inference Market & Model Routing: Commodity or Increasing in Value? | LangChain CEO
videoCrusoe CEO: Why Everyone Gets GPU Depreciation & AI Energy Costs Wrong
videoThe Untold Story of Higgsfield | Burning $4M a Month on AI Models | CEO, Alex Mashrabov
