The robotaxi growth lever the keynote skips is the operating domain. Waymo LLC's January 28, 2025 grant US12210947B2 claims a road-condition deep-learning model — inferring conditions like wet or degraded roads to adjust driving behavior.

Read the domain economics in the claim. The CPC tags — G06N 20/00 and G06N 3/08 learning, B60W 40/06 road-condition estimation, B60W 60/0015 autonomous operation, plus G05D weather-and-route classes — describe extending the conditions the system can handle safely. Every condition added expands the operational design domain.

“The technology relates to using on-board sensor data, off-board information and a deep learning model to classify road wetness and/or to perform a regression analysis on road wetness based on a set of input information.”— U.S. Patent No. 12,210,947 source

The decoder problem is that robotaxi revenue is gated by domain, not just by city. A fleet that only runs in clear weather on mapped streets serves a fraction of potential rides; one that handles rain, night, and degraded roads serves more. Each domain expansion is R&D spend that directly enlarges the addressable, serviceable market.

For a fundamentals reader, this reframes robotaxi growth as a sequence of domain unlocks, each with a cost and a revenue payoff. Waymo's expansion city by city and condition by condition is exactly this curve, and patents like this are the capitalized cost of widening the envelope.

The honest limit: a road-condition model patent does not disclose Waymo's serviceable miles, cost per domain expansion, or revenue. It establishes the mechanism for widening the domain — the relevant economic lever.

The takeaway for the money desk: robotaxi economics are operating-domain economics. Read condition-handling patents as the spend that converts a narrow demo service into a broad, revenue-generating one.