It’s a custom PC, they’re lucky it wakes from sleep at all.
Absolutely untrue, and I’ve run the numbers to prove it. Prebuilt PCs have lower TCO because they extensively test configurations and have aggressive pricing around support contracts. Every broken pc costs them! Some vendors are better than others, as well.
Now, if we’re talking about gaming PCs, the numbers are a little different there. It makes sense to roll your own for that. But for most people, building their own pc yields nothing but the entertainment of building their own pc.
With the prebuild OEMs their unifying warranty is indeed a plus, but it becomes a curse when it's finished - it is frequently not possible to replace failed parts on these systems at reasonable expense, due to customised model-specific parts (non-standard motherboard shapes, non-ATX power supplies, weird CPU Cooler mounting hardware) - in the gaming world this is alive and well today in Dell's Alienware and own brand gaming PCs, and in the workstation world, basically anything Dual Socket is frequently super custom.
The custom build may have more utility after one of it's standardised components fails out of warranty, as this standardisation allows for a choice of replacement (either in matters of cost or performance). If the prebuild fails out of warranty you may have some eWaste on your hands if a non standardised component fails.
But, still, have such glass-walled been tested for ESD/RF, if they are not sold as systems? Or are those transparent parts made of some RF-blocking material?
Surely a vendor such as HP would have the paperwork to show that their cases are tested with their motherboards, but a small system builder company (e.g. PC component seller that can also assemble the system) probably would not have done such testing..
Lets look at some PCs I've built. An AMD FX-9590 with dual 570HD GPUs in SLI. That's 700w right there. And it was impossible to keep cool, sounded like a vacuum cleaner, and felt like a space heater when you walked by. Not cost effective or efficient. Motherboard failure, but not before one of the 570s quit.
I then had a first gen Ryzen. I had so many HDDs that I had to have dual PSU's for a total of 1,850w, but not redundant like most servers are. That eventually failed due to motherboard failure. Also lots of drive failures. Way more than an OEM box. And when they failed I had to take the system offline and tear it all apart. Modern servers have hot swappable HDDs and even PSUs... so you can replace these parts without tools and without bringing the server offline.
Now lets look at a Dell R700 series server. Dual, redundant PSU's, ECC memory, redundant systems everywhere, remote management that consumer boards just can't compete with, fit and finish is perfect, drivers just work, and if it doesn't you can call someone and they will make you whole again without "troubleshooting first". When a PSU in a server fails, the server sends you an email. When the server needs a hard reboot, you can reboot it from the BMC even if the OS is panicking. There is a little screen on the front that changes colors when there is a problem, and there are color coded POST error codes.
You usually hear the argument "build your own" from hobbyists and gamers, and I do fit into that category typing this on a full tower Ryzen with a RAID 5 array for all my games. However, the broad statement of " Custom built PCs are overall more efficient, cost effective, and reliable" is simply not true. It doesn't make sense for (most) IT Departments to make, build, or custom order PCs for pretty much any reason. Even when I managed a factory network, the engineering machines were Dell Precision Workstations and they rarely ever had problems. If they were using 10x custom gaming computers I guarantee there would have been problems all the time. Engineers aren't gamers, and most aren't "custom PC" people. The tolerance for problems is very, very low. Time is money. You do a lot of your own troubleshooting that people at work doing their everyday jobs aren't going to be willing or able to do on their own.
The IT Deparment would rather a PC with 9,000 CPU marks and 2 support tickets than a machine with 12,000 CPU marks and 25 support tickets.
And that even those workstations you think are reliable can and do have problems (I have first hand experience of that), like all computers. It’s the response time and contracts that I wanted to problems that I used to look for when buying workstations and that doesn’t always help when it comes to firmware issues.
There are plenty of data centres and companies that are running what you call custom builds. Look at Hetzner, backblaze, or even bigger companies that role their own.
$100k is a drop in the ocean for experience, that is like a typical week purchasing for some, and wouldn’t even cover the cost of some machines I have used.