blog

At Earthmover, we are building the cloud platform for multidimensional array data (a.k.a. tensors). While the need for such a platform is obvious to practitioners in weather, climate, and geospatial data science (not to mention AI, where tensors reign supreme), folks from the mainstream data world often ask, why don’t you just convert everything to tabular? Indeed, in certain corners of the internet, one gets the impression that cloud data management and analytics is basically fully solved by technologies like Parquet, Arrow, DuckDB, etc. (For example, nothing about array or tensor storage appears at all on Bessemer’s Roadmap: Data 3.0 in the Lakehouse Era.) This post explains why we need Array-native data systems from first principles. It’s aimed at a technical audience familiar with m…