It started with Hadoop which was inspired by what existed at Google and became popular in enterprises all around the world who wanted a cheaper/better way to deal with their data than Oracle.
Spark came about as a solution to the complexity of Hive/Pig etc. And then once companies were able to build reliable data pipelines we started to see AI being able to be layered on top.
Data models generated by intentional human action e.g. clicking a link, sending a message, buying something, etc are universally small. There is a limit on the number of humans and the number of intentional events they can generate per second regardless of data model.
Data models generated by machines, on the other hand, can be several orders of magnitude higher velocity and higher volume, and the data model size is unbounded. These are often some of the most interesting and under-utilized data models that exist because they can get at many facts about the world that are not obtainable from the intentional human data models.