What we have done with this in the past is:
- Partition the table in SSAS / AAS - perhaps partition by week
- Update the rows in your source that SSAS / AAS loads from.
- Keep track of which weeks the changes exist in
- Process only the table partitions that are affected.
Most of the time it will be a small number of partitions involved, and you'll only average around 115000 rows per partition, which should load quickly.