The pull a large dataset exerts on the applications and services that use it, which end up running near it because moving the data is slow and costly.
Data Gravity shapes enterprise architecture. Moving large datasets to cloud-based AI models is often impractical because of bandwidth costs and latency, so organizations deploy the models where the data already lives: in On-Premises Infrastructure and Private AI deployments inside the enterprise boundary.
See also Enterprise Architecture, On-Premises, Private AI, Data Mobilization, Data Lakehouse and Vector Database.