The most important infrastructure shift in enterprise AI has nothing to do with models. It is context. At Snowflake Summit, one theme came up in every conversation: AI agents cannot operate on raw data. They need to know what the data means, how it relates to everything else, and what action should follow. That is not a model problem. That is a data layer problem. We built modern data platforms for humans. Dashboards showed humans what happened. Humans interpreted, decided, acted. That loop worked. Agents do not have that loop. They need context embedded at the source. Semantic models. Business definitions baked into the catalog layer. Not interpreted downstream — built in from the start. Snowflake Horizon Context is infrastructure for that shift. Intent-based governance that lets you describe the policy outcome and deploys it for you. IBM extended it to mainframe. Zero-copy. The last isolated data environment is now in the semantic layer. The companies still asking "where does your data live?" are behind by one full cycle. The question now is: does your data know what it means? If not, your agents are guessing. At scale. With your data.