I worked at a SaaS that had a 220 million a year bill. We did a ton to reduce the bill (reserved instances, right sizing, etc etc). 2020 rolled around and the business decided to cut back, we were able to save 20~% with being more aggressive with spot instances, moving from vpc connections via transit gateway to more peering, swapped to karpenter, etc. If there is anything I've learned is that there can always be money saved.

Related:

Figma spends $300k on AWS daily

https://news.ycombinator.com/item?id=44462016

>> what everyone’s missing is that these spend targets are post discount.

That makes it worse, not better.

>> Here’s what the pearl-clutchers don’t fully grasp: 12% of revenue on cloud infrastructure for a compute-intensive, real-time collaborative platform is completely reasonable.

The blog post author switches the discussion from cash outlay to percentage of revenue.

In every facet of your life, pay careful attention to when someone switches the numbers to percentage instead of cash ……. it has a giant impact on perception of value and if it’s happening and you don’t notice it then you’re the patsy.

Conversely, when you are negotiating, ensure you always have this tool in your toolbox to turn the negotiation in your favor, it’s an amazing way to get the upper hand whilst persuading the other party they are getting a good deal.

Masterfully used here it is clear the author is a negotiator.

I am not persuaded that a daily $300k AWS bill is OK for a company like Figma without particularly special computing requirements like AI. I am persuaded that you should engage Duckbill if you wish to convince management that your ridiculous AWS bill is normal and all is good.