I wrote below in 2022 on HN [1] when everyone was panicking about AWS growth slowing down.

* >>Amazon said Thursday that revenue growth in its cloud-computing unit slowed in the third quarter to 27.5%.*

27.5%. It is lower that their previous 33% growth over the past few years, but at the current size of AWS growing 27.5% is still ridiculously good. To put this in perspective, if AWS continues to grow at 33% in 2022 and 2023. Then the whole 2023 33% growth alone, would equal to the size of the entire AWS in 2018. It is not the first time Amazon said they are limited by how fast they are building out Datacenter and getting hardware resources ready.

That was in 2022. They nearly double their 2018 size alone in a single year.

I don't understand back then. I still can't get my head around it now. With or without AI. With AI the number and scale just grows beyond my imagination. CPU power per socket or per Rack have increased every single year. What used to take 10 racks could now be replaced by 1. I would have expected slowly replacing old Rack to newer ones would have been enough with slower Datacenter growth. That is not to mention software have gotten faster and efficient over the years. JVM, PHP, Ruby, C, Database etc over the past 10 - 15 years.

Instead we keep growing, not only that; AI have shown they seems to have infinite appetite for computing resources. I know this is classic Jevons Paradox but the scale [2]. It is mind boggling numbers.

[1] https://news.ycombinator.com/item?id=33384628

[2] I remember the last time I had scale issues was I can't compute in my head how Apple will be a trillion dollar company by 2020. That was written on Appleinsider in ~2012. We now have multiple trillions dollar companies. The TAM of some of these market continue to amaze me.

AI's appetite for computation scales with the willingness of their funders to provide heavily discounted tokens to the general public.

The basic play here is to get companies to fire people and adapt internal processes to the current heavily subsidized AI farms, and then jack the rates once the switching costs become untenable ... particularly if they can get a huge percentage of human programmers to quit the industry.

This is effectively "dumping" in the economic sense.