When farming, disk speed is not critical, just size. Thus the idea of recycling millions of HDDs in a low power scenario. A few seeks here or there on your old, unused terabytes, but you need to have enough space to offer, otherwise your chances are zero. Today's farming is dominated by pools, some using non-official, closed source binaries of unknown origin. Official pooling is, supposedly, coming soon.
Plotting, which is preparing the files that are farmed, is, as of today, extremely IO intensive, but still cheaper, energy-wise, than mining. A single 100GB plot (you need ~20TB to have any chances of farming success) can take hours to plot. So people are resorting to extremely fast SSDs, or NVMe in RAID0 in PCIe 8x/16x cards to slash that time. And, in the process, wearing them out. Can't find the link now, but some were saying normal SSDs are ruined after 6 weeks of 24/7. Plotting keeps improving, although I'm not sure if it's got a lower performance bound by design.
Now with this new currency, you could maybe even run cloud storage. Maybe we are accidentally building the futur alternatives to Amazon cloud services.
The chips that are mining bitcoin are not general purpose computers.
This is a part of why NVidia cracked down on gpu mining recently - when ASIC mining became more profitable than GPUs, the secondhand market got flooded with cheap high quality cards (thus ruining the primary sale market for new GPUs) and people flooded NVidia support with support queries for slow or unreliable GPUs.