SpaceXAI’s Biggest Business May Soon Be AI
Overall Thesis
SpaceX's AI division is rapidly advancing with Grok 4.6 achieving frontier-level performance at lower cost, and Elon projects AI revenue could surpass other SpaceX revenue by September.
Narratives
The hosts discuss a deepening, unofficial partnership between Nvidia and Elon Musk's companies, framing it as a 'virtual merger' built around chip design and manufacturing rather than corporate structure. They note that Grok 4.6 was trained and run on Nvidia's GB300 chips, and Nvidia publicly congratulated the SpaceX AI team on the release, which the guest sees as evidence of a much closer collaborative relationship.
Key Arguments
- Grok 4.6 runs on Nvidia GB300 chips and Nvidia congratulated the SpaceX AI team on its release
- Guest argues Nvidia and SpaceX/Tesla are forming a 'virtual merger' at the product level via shared chip design (Terrafab) rather than a corporate merger
- Nvidia's massive compute financing could partly flow into SpaceX's data center buildout
The discussion frames Elon Musk's overall AI strategy as centered on engineering and physical-world AI (vision, robotics, physics) rather than pure language models, which the guest sees as differentiating Tesla/SpaceX from OpenAI and Anthropic. The hosts speculate that future investment value may come from products like robotaxis, Optimus, and AI infrastructure rather than traditional corporate structures, but no specific Tesla stock price target is given.
Key Arguments
- Elon's strategy is described as focused on engineering and physical AI (vision, physics-accurate simulation) rather than just LLMs
- Guest suggests future investing will be in products like robotaxi and Optimus rather than in the corporation itself
- Deep Nvidia collaboration seen as reinforcing Tesla/SpaceX's AI infrastructure advantage
Hedges & Caveats
- Elon's claims about future model performance (Grok 4.7) are projections, not guarantees
- Revenue projections for AI surpassing other SpaceX business by September are Elon's claims without independent verification
- Comparisons to competitors (Anthropic, OpenAI) are based on benchmark tests which may not reflect real-world performance
- Early user impressions of Grockbot are anecdotal