It seems like there will be a lot of hardware that gets discarded every couple of years as new, more powerful systems are made available.

AWS put out a video and article on how it recycles a large amount of its hardware, since it is built for maintainability and repairability.

How true is that? Does it apply to other datacenter operators as well?

What will happen to all the parallel compute cards that will get upgraded soon? They can't be reused as GPU's for gamers, can they?

Generally large companies will have an existing vendor that they use to dispose their IT equipment through. They will shred parts like storage devices and anything that can reasonably be resold will be sold through various auction houses.

Sticks of ram will certainly be resold, custom aws motherboards - not so much.

I have seen custom (unpublished) intel cpu parts on ebay before which are almost certainly aws's custom ones.

Almost nothing will get used by consumers - enterprise server gear is designed for heat/air speed/noise/energy cost requirements which are incompatible with consumer requirements. It's recycled only in the sense that a smaller business might be interested in it because at the end of its economic life its now cheap to buy (but not cheap to run).

https://cloudninjas.com/

https://savemyserver.com/

https://unixsurplus.com/

In Orlando, if you go to disney and you are a techie, it is a must see, is a company called SkyCraft Surplus. They go to companies who are getting rid of old equipment and resell it. Some stuff is new, much of it unused. Brought a friend there (EE major) and he was fascinated by the fact that they had a lot of old Radio Shack chips, diodes etc.

Tell them I sent you and they will ... well nothing, but I am happy to support them, as it brings me much joy whenever I visit.

It depends on what it is. A lot of these places don't own any of the hardware they just lease it. When the lease is up is when they cycle it out. Then it goes to resellers and often ends up on eBay or bulk sold to lower tier data centers. Depending on what it is maybe even shipped to other countries.

Most of it goes through ITAD channels: storage gets wiped or shredded, reusable gear gets resold, and the rest is recycled by weight. Accelerator cards usually get a second life with smaller clouds/labs, but older SXM generations fall off fast. The useful metric I wish providers published is split by class: reused vs recycled vs trashed.

What will happen to all the parallel compute cards that will get upgraded soon? They can't be reused as GPU's for gamers, can they?

You're right; V100/A100/H100 "GPUs" do not have the hardware to display graphics and they generally require custom SXM motherboards. Most of them will sit on eBay for a while and then be scrapped when no one buys them.

Most of this HW is at the end of its life. It's not about time but usage. And data centers use the HW to the limit.

huge ITAD/ITAM companies. G / AWS mega datacenters have their own processes but ITAD companies will then upgrade, recycle, or straight up sell devices or full machines. they of course, have government standards when wiping hard drives / dram etc.