Latest / Minds, Bodies, and Terawatts / Tesla's Terafab Chip Facility Targets AI5 Chip Production, with Implications for Intel Foundry Deal

Transcript

Tesla's Terafab Chip Facility Targets AI5 Chip Production, with Implications for Intel Foundry Deal · Minds, Bodies, and Terawatts

• 0:00
  [Host] This is Minds, Bodies, and Terawatts, where the silicon meets the road. SpaceX filed paperwork on May sixth disclosing a fifty-five billion dollar first phase for Terafab, its chip factory in Grimes County, Texas. Total investment across all phases could reach a hundred and nineteen billion.
• 0:20
  [Guest] To put that in perspective, TSMC spent about thirty billion on its Arizona fabs. SpaceX is committing nearly double that for Phase 1 alone. The long-term target is one million wafer starts per month, producing custom AI and memory chips for Tesla vehicles, Optimus robots, and SpaceX orbital computing.
• 0:41
  [Host] The surprise in this story is Intel. Wasn't the assumption that Terafab shuts Intel out?
• 0:47
  [Guest] The opposite happened. Intel joined Terafab on April seventh as its manufacturing partner. Musk confirmed on Tesla's Q1 earnings call that Terafab will use Intel's 14A process. Intel's stock more than doubled in April, its best month on record, largely on this news.
• 1:06
  [Host] So Wedbush's read is bullish for Intel?
• 1:09
  [Guest] Dan Ives called the project significant for Intel and flagged ASML and Applied Materials as beneficiaries. But Wedbush was equally blunt about risk. The gap between a filing in rural Texas and an operating fab is wide. No project at this scale has been linear.
• 1:29
  [Host] That's the pattern with Musk, though. Cybertruck shipped two years late. Full Self-Driving has been next year since 2016. But the Cybertruck exists and Starship flies.
• 1:40
  [Guest] Consistently wrong on when, right on whether. Morgan Stanley estimated meaningful chip output no earlier than mid-2028. Even at a fraction of the claimed capacity, the facility would reshape global semiconductor supply.
• 1:54
  [Host] Here's what bothers me. When Bell Labs invented the transistor, knowledge diffused outward. The whole electronics industry benefited. Terafab's output stays locked inside one ecosystem.
• 2:05
  [Guest] Tesla chips go in Tesla cars. Those cars train xAI models. Those models run Optimus robots built in Tesla factories. The AI5 chip handles motor control on-device, no cloud round-trip. Every link feeds the next. Nothing leaves the loop.
• 2:22
  [Host] And SpaceX merged with xAI earlier this year in a deal valuing the combined entity at one-point-seven-five trillion dollars. One man now controls the rockets, the AI, the robots, and the silicon underneath all of it.
• 2:37
  [Guest] That's the concentration alarm. There is no regulatory framework for a vertically integrated chip-to-deployment pipeline. Nobody audits what gets baked into that silicon before it ships in millions of vehicles.
• 2:51
  [Host] Intel choosing to enter that orbit rather than compete with it might be the more telling signal. Even a legacy chipmaker decided the gravitational pull was too strong to resist.
• 3:04
  [Guest] Which is exactly the mechanism Robert Michels described. Power doesn't just concentrate through force. It concentrates because joining the winning structure looks rational for every individual actor, even as it erodes collective oversight.
• 3:18
  [Host] Patrick Deglon has a law for this in Unscarcity. Power Must Decay, he writes. All authority must be temporary, or it becomes tyranny. Musk's architecture has no decay function. No term limit, no sunset clause, no external audit at any layer.
• 3:36
  [Guest] And the distance between what this technology does and what institutions can govern widens with every wafer that rolls off that line. For more on how concentrated power reshapes the AI landscape, head to unscarcity dot ai. The book is on Amazon, the audiobook on Spotify. This has been Minds, Bodies, and Terawatts.