Software engineer Dave McCrory coined the term in 2010, comparing data to a mass that pulls other things toward it. As a dataset grows, teams build applications, pipelines, and analytics close to it, because moving large volumes across networks or clouds costs time and money and adds latency. Those applications create more data in the same place, which strengthens the pull.
Data gravity shapes decisions about where workloads run. A company with petabytes of history in one cloud data warehouse often trains and serves models in that environment instead of copying the data elsewhere. Rules such as data residency can also fix data to a region. For AI projects, teams often bring compute and data preparation to where the data already sits, and they budget for transfer costs when a task needs data from several places.