Elon's New Factory Makes NVIDIA Look Small
Overall Thesis
Tesla and SpaceX's Terafab mega-factory in Texas will vertically integrate chip design, fabrication, and memory production to address the massive supply-demand gap in AI compute, positioning Tesla stock as a beneficiary of in-house chip supply.
Narratives
The host frames Tesla's jointly-owned Terafab chip project with SpaceX as a way for Tesla to secure its own supply of AI5/AI6 chips for FSD and Optimus, removing dependency on external foundries like TSMC and Samsung. He argues this addresses the single biggest bottleneck risk in the Optimus ramp, even while flagging that headline figures like the 'terawatt' output target and 'worth more than TSMC+Micron+Nvidia' claims are unproven ambition rather than signed fact.
Key Arguments
- Terafab is jointly owned by Tesla and SpaceX, giving Tesla in-house control over AI5/AI6 chip supply for FSD and Optimus
- 16.8 billion already committed and construction/permits are real and underway at Giga Texas and Grimes County
- Removing reliance on Samsung/TSMC foundry capacity addresses the scariest single point of failure in the Optimus ramp
- Elon's rough split allocates about 25% of Terafab output to Tesla Optimus
Predictions (1)
Micron is referenced as the memory-making counterpart that Terafab aims to replicate in-house, but the host makes no direct claims about Micron's stock or business outlook beyond its role as an industry benchmark.
Key Arguments
- Micron represents the memory-fabrication step that Terafab is designed to internalize
- Global supply of advanced chips and memory is described as growing only ~20% a year versus ~200% AI compute demand growth
Nvidia is described as the current design leader for AI chips and a continuing supplier to SpaceX's data centers and the new Star Mind satellite compute partnership. The host notes Nvidia 'isn't going anywhere soon' even as Tesla builds Terafab to reduce long-term dependency on outside chipmakers.
Key Arguments
- Nvidia is the 'king of design' for AI chips and remains Terafab's conceptual analog for the design step
- SpaceX's data centers run entirely on Nvidia hardware today
- SpaceX and Nvidia are partnering on the Star Mind AI1 satellite compute payload using Nvidia Rubin GPUs and Vera CPUs
TSMC is cited as the fabrication leader whose 'whole empire' Terafab seeks to replicate, with the host noting TSMC's competitive moat took roughly 30 years of process refinement to build, implying skepticism that Tesla can quickly replicate that edge.
Key Arguments
- TSMC's fabrication lead exists because of decades of accumulated process detail that can't simply be bought with capital
- TSMC continues to supply Tesla's chips while Terafab ramps up
Hedges & Caveats
- Terafab's one terawatt annual production target is described as a 'stage number' with nothing shipped yet
- Project execution risk acknowledged - Elon's mega-projects have history of delays
- Dependent on successful ramp of manufacturing at unprecedented scale
- Competitive response from TSMC, Samsung, and Micron not fully addressed
- First chips promised 'next year' but timeline history suggests caution warranted