1 min read

LLM Economics, LLC

Hypothesis: The unit economics of the machine learning space LLM 2020's startups are not similar to the social networking space in the mid-2000s. This will create a five year vacuum in VC's expecting returns


  • Metcalfe's law is at play for people-run networks, but interfacing with LLM is a single player game
  • Computers are bought for their GPU number crunching ability, not their ability to serve data to millions of users instantly
  • Most of the cost is upfront during training period
  • Correctly built AI companies will slowly/quickly drive out middle management related work, decreasing the need to hire employees to scale/build/market a project