It really depends on which workloads. We have teams working only on FinOps to optimize our architecture to take advantage of various cloud pricing models.
Enterprise spend on cloud is a massive topic right now and definitely in the spotlight. A consensus is developing that the current implementation of public cloud promotes lock-in and overall much more expensive at scale.
Cloud is complex and engineers are expensive. Instead of throwing CPU cycles and disk at the problem (which is cheap compared to engineer time), we are nickel-and-dimeing everything, and changing our architectures to hack the cloud pricing models. It's pretty silly.
