The dual-track roadmap splits Tesla's silicon ambitions into two lanes. SI6 handles inference on the ground; Dojo D3 handles training and, per Tesla, can eventually go to orbital data centers. The two chips serve different workloads and different environments, and the company is presenting them as a single strategy.

The branding shift from AI to SI — "Super Intelligence" — repositions the chip line as a platform rather than a product. The headline number is the 80x gap between AI4 and SI6. That figure is what will determine whether Tesla's inference stack keeps pace with the model sizes it trains on Dojo.

The orbital claim is the more speculative half of the roadmap. "Can eventually go" is not a deployment schedule; it is a design intent with no attached timeline or partner.

Sources: Tom's Hardware, dev.to. The hole: the 80x figure is Tesla's own claim with no independent benchmark, and the orbital deployment for Dojo D3 is aspirational, not a confirmed plan.

The packet contains two claims from a single disclosure event. The 80x figure is a vendor claim with no third-party benchmark attached. The orbital deployment language is explicitly conditional — "can eventually go" — which means it is a design capability, not a committed deployment. Both numbers are therefore unverified in different ways: one is unmeasured, the other is unscheduled. The dual-track framing (inference on the ground, training in orbit) is the structural move, but the structural move rests on two numbers that have not been independently confirmed.