Waymo, Zoox, and Tesla: Operational Implications of Self‑Driving Cars (gadallon.substack.com)

Three different robotaxi strategies are now being tested in real cities: Waymo’s retrofit, multi‑sensor fleet; Zoox’s purpose‑built bidirectional vehicle; and Tesla’s camera‑only, retrofit Model Y pilot (with a planned CyberCab). Waymo’s 6th‑gen stack uses redundancy at scale (13 cameras, 4 lidars, 6 radars, multiple radars and actuators, high‑res maps) and reports big safety gains (e.g., ~91% fewer severe‑injury crashes vs. human drivers). Zoox’s toaster‑shaped taxi integrates overlapping lidar/radar/camera arrays, four‑wheel steering, redundant power/steering and long‑runtime batteries, but has logged a few minor incidents during early service. Tesla relies on “Tesla Vision” (no lidar/radar), safety monitors in pilot vehicles, and a low‑cost, owner‑centric scaling vision; its Austin pilot recorded three crashes in ~7,000 miles and drew an NHTSA probe. The operational implications are stark: multi‑sensor, purpose‑built designs prioritize robustness, redundancy and predictable safety at higher capital and operating cost, favoring operator‑owned fleets (Waymo, Zoox/Amazon). Tesla’s camera‑only and retrofit path lowers hardware costs and enables faster scale or peer‑to‑peer models, but raises questions about all‑weather reliability, regulatory compliance and early safety performance. Ultimately, the race balances cost and speed-to-scale against redundancy and real‑world robustness; winners will be determined by who can prove low incident rates, sustainable unit economics, and regulatory acceptance.