I was actually surprised by how cheap it is right now; I expected more than 300. I was thinking of building a new PC right when the ram crisis started 
I’m sure, when the bubble will burst (when, not if), there will be a lot cheap and unused 2nd hand hardware on the market.
1 Like
Yep, that’s a huge problem. That’s one I’m struggle for decades with. With the GPU I have at home, with the projects I have done with classic rendering in jobs. 10+ PCs running for days or weeks to render videos. Digital “Art” will consume electricity. Same with development in Godot. The problem is, that there are now many millions who “render” (generate) videos, not only some 3D nerds. Same with code.
I hope it will get better with TPUs instead of GPUs.
And a positive aspect of AI are denoiser and upscaler. Most 3D artists are using them now for realtime or rendering. They reduce the computation and energy consumption.
2 Likes
I fear that’s part of the smoke and mirrors : locally it seems to use less power and computing time, but in the AI data center, every single task done through ML and LLMs is more costly in both electricity and computing power than done traditionally.
To that, you have to add the exploitative enshittified data labelling work done by humans in developing economies to that cost.
So no, nothing in what we currently call AI leads to economies of scale, quite the contrary: look at SSDs (two of the major bulk vendors are out of stock for 2026) and RAM prices. Let’s not forget consumer GPUs which are artificially inflated, as they’re pretty much useless for AI data centers.
Cheers, have fun making games!
2 Likes
The denoiser (and upscaler) are reducing the energy consumption. If you only need to render <10 samples instead of 1000+ it’s a big difference. (And it’s generated on your GPU).
Example:
Gross, it’s disgusting that I receive emails about slop, where do I unsubscribe?
As a small tangent, I’m annoyed by Blender hype as well. Every new feature “blows minds” and “changes everything”.