Tesla unveils AI accelerator roadmap, adopts new 'SI' branding in 'Super Intelligence' push

Tesla's move is not about making chips—it's about building a vertical compute-sovereignty chain from silicon to orbit. The Dojo D3/SI6 dual-chip matrix directly mirrors NVIDIA's H100+B200 tier, but the critical differentiator is binding inference silicon to autonomous driving and humanoid robotics, creating a data-flywheel→chip-iteration→model-evolution loop that pure-play chip vendors cannot replicate. On the supply side, advanced packaging remains locked to CoWoS capacity in Taiwan, China. Any tightening of U.S. export controls would immediately disrupt Tesla's production cadence. The orbital data-center concept is more radical still: if realized, it bypasses terrestrial power and thermal constraints, yet launch costs, in-orbit maintenance, and spectrum licensing create regulatory barriers unlikely to clear within 12 months. NVIDIA's most probable counter: accelerate Blackwell Ultra ecosystem lock-in while offering custom ASIC services to automakers, dismantling Tesla's 'in-house equals moat' narrative. The 24-month tail effect: if the 'SI' brand successfully extends from silicon to the model layer, Tesla's valuation anchor shifts from auto P/E to AI-platform P/S—rewriting its market-cap ceiling entirely.