This episode of Equity discusses OpenAI's custom chip 'Jalapeño' built with Broadcom to reduce inference costs and hedge against Nvidia dominance. The hosts also explore AI agents running on endless 'loops', a robotics SPAC from Agility Robotics, and A24's AI deal with Google DeepMind, highlighting themes of cost control, compute constraints, and industry self-interest.
Summarized by Podsumo
OpenAI's custom chip 'Jalapeño' with Broadcom focuses on inference to lower costs and reduce Nvidia dependency, similar to Apple's shift from Intel.
AI 'loops' (agents running endlessly) are the new hype, but hosts are skeptical for consumer use due to token costs and complexity.
Agility Robotics is going public via SPAC at $2.5 billion, a rare mature company with real revenue and customer sites like Toyota.
A24's investment from Google DeepMind for AI tools signals Hollywood's cautious embrace of AI, despite potential filmmaker backlash.
Memory chip crunch is driving up costs, affecting Apple prices and potentially ChatGPT pricing.
"The lowering cost piece is really important because it isn't just dealing with... the vast world of chips. — Kirsten Korosek"
"I think about... Apple breaking with Intel... they took what was at the time maybe arguably like sort of many middling products... and turned them into things that were like incredibly desirable. — Sean O'Kane"
"Loops are the new term du jour... AI agents that are now running endlessly in the background, like on a loop, and prompting other agents. — Anthony Ha"