The camera-versus-LiDAR debate is the surface. The deeper split is visible only when you read two filings together, both retrievable via EdgarBeast. Nvidia's FY2025 10-K, filed at sec.gov, says the company delivers a complete end-to-end solution for the AV market under the DRIVE Hyperion brand, built on its automotive relationships.
Tesla's FY2025 10-K takes the opposite posture: it applies the AI learnings from its own self-driving technology to its own products, including Optimus, on its own custom inference chips (sec.gov). Nvidia wants to power everyone's autonomy; Tesla wants to own its own end to end. That is a business-model fight wearing a sensor-debate costume.
The two strategies have different failure modes, and the filings imply them. Nvidia's horizontal bet wins if autonomy fragments across many automakers who would rather buy a platform than build one — its risk is a major customer in-sourcing the stack. Tesla's vertical bet wins if owning the full loop produces a data-and-iteration advantage — its risk is carrying the entire R&D burden alone.
Sensor philosophy follows from the model, not the other way around. A platform supplier benefits from sensor-agnosticism, so Nvidia's reference stack accommodates many sensor suites. A vertically integrated player can make a single opinionated bet — Tesla's reuse of one perception stack across car and robot only works because it controls the whole pipeline and standardizes the sensing.
Treat both camps fairly and the synthesis is this: there is no single 'right' autonomy architecture in these filings, only two coherent business models with different bets about how the market consolidates. Read sec.gov and Tesla's 10-K together and the vision-versus-mapped argument finally sits where it belongs — downstream of strategy. Both filings surfaced via EdgarBeast.