In the Physical AI era, ADAS is evolving from modular, rule-heavy pipelines toward more integrated, learning-centric stacks that close the loop from perception to action. This talk explains how that shift is reshaping vehicle architectures—driving higher compute intensity, new domain controller/platform choices, and updated sensor requirements across cameras, radar (including imaging radar), and selective LiDAR. Beyond technology, Physical AI is also redrawing the automotive supply chain. As OEMs bring more software and data ownership in-house, and SoC vendors ship richer SDKs and reference stacks, Tier-1s face “platformization” pressure and must respond with stronger system-level bundles and partnerships. We conclude with a market outlook, highlighting where sensor and compute content per vehicle is likely to grow.
