bottleneckScore 75/100Research

Hotz: Google Cloud ToS bars training autonomous vehicles on TPUs, leaving Nvidia a training-compute monopoly

George Hotz· Comma.ai· Robotics· —· pub 2026-08-20· about Google (TPU) / NVIDIA (NVDA)
“Your options are Nvidia or Google... Google's terms of service said explicitly you are not allowed to use Google Cloud ML for training autonomous vehicles or for doing anything that competes with Google without Google's prior written permission... So that leaves you with the monopoly — Nvidia.”

Why it matters

A first-party AI practitioner states that Google's cloud ToS effectively blocks TPU use for competing workloads, making Nvidia the only viable training platform for non-hyperscaler AI companies.

Investment implication

Reinforces Nvidia's moat in AI training silicon among independent AI/autonomy startups, while flagging that Google's restrictive TPU access limits its cloud AI competitive position versus AWS and Azure.

Source

George Hotz: Simulation, comma.ai, Openpilot, GPT-3 & AGI | Lex #132 (YouTube)
← All signals