Otherwise, there would be market for people to sell their spare CPU cycles.
The only entities selling CPU cycles are big cloud providers who have much higher utilization in general.
Your return on the opportunity cost is maximized if you can do work closest to the time the card was acquired.
If you buy a card and leave it in the box for 10 years you have not 'consumed' the card, but you have wasted a few hundred dollars.
From an electricity and financial perspective if you have valuable work to do, your costs are minimized if you do that work closest to the purchase date.
This is obvious if you have a significant workload, but maybe not as obvious if you are running CAD. Is the cost of a new card worth the time saved by the new card? Significantly more efficient if you can keep it busy.
The other question to ask, if you compile programs, write webpages, and edit photoshop images: are you concerned with hardware degradation with your CPU if you run at 100% all time time, for example, compiling chrome that takes 3+ hours, applying photoshop filters to an image to get a production quality output, or for rendering a 3 hour long animated movie?
Do you worry about the GPU when you are retraining a GPT-2 or GPT-3 scale AI framework?
These questions are kind of ... dependent on each person. Presumably you buy a computer in order to perform computations. That is its purpose.
Just because it's crunching bitcoins/cryptocurrencies doesn't degrade it any more than ... say running an AI framework on a GPU for 3 months continuously to generate a self driving model ...