Agent gravity is the defining platform battle of this decade. The same way data gravity locked workloads into platforms during the 2010s, agent gravity will determine who controls compute, pipelines, and revenue in the 2020s. The more agents and data running through a platform, the harder it becomes to leave.
The Databricks-Microsoft conflict makes this concrete. A new Databricks feature inside Microsoft's ecosystem allowed Power BI customers to build AI agents and manage data inside Databricks, pulling workloads away from Microsoft Fabric, its competing data product. Agents built on Databricks can tap the semantic layer, migrate data to other cloud warehouses, and publish to rival BI systems. Users, often without realizing it, shift profitable workloads off the incumbent platform.
The full piece by Tomasz Tunguz is worth reading for how it maps agent gravity as a strategic force, not just a product anecdote. The framework it builds around who controls the agent runtime, and therefore controls data and compute decisions, has direct implications for every enterprise platform decision being made right now.
[READ ORIGINAL →]