The one paragraph Google’s CFO cannot put in a 10-Q

Generative AI materially alters our cost structure in ways that are not yet reflected in reported unit economics. Unlike traditional search, which benefits from near-zero marginal cost per query, large-scale AI inference incurs materially higher variable compute costs that scale with usage, verbosity, and reasoning depth rather than impressions or clicks. At present, there is no proven mechanism to monetize AI responses at a rate commensurate with their operating cost without degrading user trust, engagement, or regulatory posture. While we expect ongoing efficiency improvements in hardware and model optimization, these gains may be offset by rising usage expectations and competitive pressure to deliver more compute-intensive outputs. As a result, AI-driven products may dilute margins and cannibalize existing revenue streams before a sustainable, physics-constrained business model is established, and there can be no assurance that such a model will emerge on a timeline aligned with current capital investment levels.

– Published Friday, January 16, 2026