Layer 8: Waymo and Physical AI — Why Google’s 8-Layer Stack Compounds Into Autonomous Vehicles

This analysis is part of Google: The Gravitational Center of the AI Market, a deep dive by The Business Engineer.

Others call it a moonshot. Google treats it as the next distribution surface.

Waymo — The Only Commercially Scaled Autonomous Fleet

Expansion across 15+ cities targeted by end of 2025: SF, LA, Phoenix, Austin, Atlanta, Miami (live and expanding) plus Tokyo (planned). Millions of paid rides. 4x trip volume growth YoY. Partnerships with Uber, Hyundai, and Toyota. Gemini powers the driving model — TPU → Gemini → Waymo. Same stack.

DeepMind Robotics — From Digital to Physical AI

Foundation: RT-2 + Gemini (vision-language-action model for robotics). Gemini Robotics: Multimodal reasoning for physical interaction — same Gemini that powers Search and Cloud. Physical Deployment: Warehouses, hospitals, factories, roads. Waymo (roads), Intrinsic (industrial), Everyday Robots heritage. Multiple physical domains, one AI stack.

Competitive Landscape

Google/Alphabet: Full stack (Waymo + DeepMind + TPU + Gemini + Maps + Street View). Silicon → model → vehicle → city. Tesla: FSD supervised, Optimus robot prototype — no fully autonomous commercial service. Amazon/Zoox: Custom vehicle, small-scale testing. Cruise/GM: Paused and restructured. Apple — as explored in the interface layer wars reshaping consumer tech — : Project Titan cancelled ($10B+ spent).

Waymo doesn’t just use Google’s AI. It is built on every layer of Google’s stack. No competitor can assemble this from parts.

Read the full analysis on The Business Engineer →