Comment by phil21

It’s valuable, it’s it’s the least valuable part of the value chain. A rounding error that looks good due to the immense scale of the numbers involved. Buys you some time.

Cheap energy and delivery is many times more valuable than spending that same amount of money on efficiency. One has your engineers and builders working on solving new problems - the other is just taking existing solved problems and making them more efficient.

I’d much rather be finding new and exciting ways of burning ridiculously clean and cheap megawatts of electricity than spending the same human capital on how to solve the same solved problem marginally more efficiently.

For the most part the low hanging fruit has been long picked and we have long past hit the point of diminishing returns.

Ideally you do both of course.

You can have your devs spend a year making your code run 50% more efficiently. Or you can use those same devs to make new features and simply buy 50% more cloud compute capacity. Computers are cheap and human capital is limited.

It’s interesting HN generally goes one direction here when it’s “their” economic interests on the line and the other way around when electric generation and grid capacity is the topic.