BigQuery data is stored (I assume) in column oriented files with indices, so a typical query reads only a tiny fraction of the stored data.