Rivian CEO: Tesla is going all in on cameras, but self-driving cars still need LiDAR (www.businessinsider.com)

Rivian CEO RJ Scaringe publicly pushed back on Elon Musk’s vision-only approach to autonomous driving, saying LiDAR remains “definitely beneficial” and could be part of a multi-sensor stack. Speaking on The Verge’s Decoder podcast, Scaringe argued that LiDAR has become much cheaper (down from “tens of thousands” to “a couple hundred bucks”) and “can do things that cameras can’t,” and that modern perception models now ingest multimodal inputs up-front rather than processing sensors in siloed pipelines. He said Rivian’s priority is rapidly building a foundation model that benefits from the maximum amount of information, and he wouldn’t rule out including LiDAR alongside cameras and radar. The exchange highlights a core technical schism in autonomy strategy: Tesla’s camera-only, vision-first approach versus multi-sensor fusion championed by companies like Waymo, Rivian and Ford. Proponents argue LiDAR improves robustness in challenging conditions (e.g., bright sunlight, low contrast) and reduces ambiguity where cameras struggle, while critics like Musk warn of “sensor contention” and added complexity. The implications are practical and architectural — choices affect sensor cost, training data, model design (uni-modal vision models vs. multimodal foundation models), compute needs, and operational safety envelopes. The debate signals that sensor selection remains an active design trade-off with real consequences for deployment timelines and reliability.