Google is your friend. Do lots of searches using multiple search terms and search engines. Do your homework based on replies you get here.
.
Since you are planning to buy new hardware, wait until after you decide on the DBMS. Sometimes software works better with specific hardware configurations.
.
In general, go with more RAM.
Go with SSD drive(s).
Here is an EXTREME example of insane specs on server hardware:
https://www.youtube.com/watch?v=9MXhGABtuCA- "fastest server" 2020 12 28 Absolutely INSANE specs (ie 1400 WATT power supply, 200watts of cooling fans, 59TB windows boot drive ... )
This "LIQID" server has absolutely insane specs. The specs they mention in the video had me ROTFLMAO! A hardware nerds "wet dream". It has "dynamic" hardware technology that I have not seen since "Tandem" mainframe hardware. It allows you to hot swap elements and dynamically reassign hardware elements between CPU boxes. Cool stuff. Unfortunately the LIQUID site makes you request price quotes. After watching the video I really could use a real world price "cold shower". I think it is safe to guess that the prices on this server hardware will make passing a kidney stone seem like a stroll in the park. And I think it is safe to say this server is overkill for your needs, price and spec wise, but wow what an insane
.
I haven't seen any reviews about creating arrays with SSD's, whether it is faster or not.
https://www.youtube.com/watch?v=N0qtu5NXhuQ - 2015 example 20tb SSD RAID
https://helpdeskgeek.com/reviews/hdd-raid-vs-ssd-raid-the-major-differences-you-should-know/
https://www.bhphotovideo.com/c/buy/ssd-arrays/ci/30905 - example of SSD RAID units sold, prices (ouch!)
(I used to have a link to an absolutely absurd demonstration of RAIDed SSDs back when they first came out. It one of those "extreme" build examples.)
https://www.youtube.com/watch?v=lzzavO5a4OQ 2017 video
.
Check to see if your DBMS can take advantage of GPU. A quick google identifies there are GPU DBMS's. The massive parallelism in GPU means that these DMBS's are intended for massive databases (orders of magnitude bigger than your estimate) and for "real time" / Online processing.
https://sqream.com/blog/different-gpu-databases/
https://medium.com/@sammedkagi/gpu-accelerated-databases-459b0e476a6e
https://www.omnisci.com/learn/gpu-database
https://www.infoworld.com/article/3327561/what-a-gpu-powered-database-can-do-for-you.html
.
I suspect that with some research you'll find that you don't need a standalone GPU on your new computer if it is dedicated as a DBMS server.
.