On one hand it's allows to shift the blame but on other hand is shows a disadvantage of hyper centralization - if AWS is down too many important services are down at the same time which makes it worse. E. g. when AWS is down it's important to have communication/monitoring services UP so engineers can discuss / co-ordinate workarounds and have good visibility but Atlassian was (is) significantly degraded today too.