Google’s State of AI Infrastructure report identifies a core bottleneck: organizations struggle to give AI agents business context and semantic meaning, not because model capabilities are lacking but because access to context is fragmented across legacy systems. According to the report, 83% of organizations believe they need infrastructure upgrades to support production agentic AI.

To address this, Google introduced the Agentic Data Cloud at Google Cloud Next 2026. It unifies data, AI models, and operational databases into a single System of Action built on an AI-native foundation—from chip to model—designed to handle agentic load patterns where a single prompt triggers independent browsing, querying, and execution across multiple systems.

The Agentic Data Cloud relies on a borderless Lakehouse running on open standards like Apache Spark and Apache Iceberg. By using native engines such as BigQuery and Spanner over these open standards, agents can read, reason over, and activate data across environments as if it were local, avoiding the latency and cost of traditional data movement.

The report highlights three major infrastructure gaps: 43% of IT leaders cite difficulty integrating with legacy APIs and data sources; 81% call out operational complexity and engineering overhead as top unforeseen scaling costs; and 36% report a lack of specialized, high-throughput vector databases for AI model grounding. The Agentic Data Cloud aims to reduce these bottlenecks through vertical integration, enabling agents to connect real-time data across analytical and operational sources with fewer network hops and better-integrated tooling.

In practice, the system uses Knowledge Catalog to aggregate and enrich data in data lakes, automatically extracting meaning from unstructured data and generating semantics to act as an active reasoning layer. This catalog gives agents the long-term memory needed to recall past user preferences while executing complex tasks, rather than reprocessing data for every query.

The report concludes that the winners in the agentic era will be those who can feed agents the right knowledge securely, cost-effectively, and at scale. The path forward requires a connected, active data ecosystem supported by infrastructure that can handle agentic demands, moving beyond storage-only systems to proactive systems of action.