Part 2/15:
Cutress begins by assessing Tesla’s Dojo system, describing it as more of a failure than a success as of now. He highlights that Dojo was conceived as an infrastructure project—an array of interconnected wafers and support silicon, rather than a single chip, designed to serve Tesla's internal needs, particularly video data labeling for AI training.
He notes that Tesla’s approach was massively ambitious, building raw infrastructure rather than a consumer product, aiming to optimize their self-driving AI training. However, a series of setbacks and industry observations cast doubt on its current effectiveness: