Persistent storage doesn't have to be powered on all the time, though, which makes the environmental impact a known thing. Factor in the cost of building + populating the drives along with having to replace them periodically and you know the cost of that storage. In comparison, PoW's future cost can skyrocket as the network expands and the difficulty goes up.

Storing the whole chain takes a bunch of space anyway, so you have a similar storage problem there too.

If you want to see how cheap persistent storage can get, look up the cost of Amazon Glacier. It's shockingly cheap - $0.005 per GB or even less.

It's a known quantity, sure, but IIRC hard drive manufacture was not necessarily a very "clean" process.

Right, but ultimately the same is true for manufacturing GPUs and ASICs. It's a problem you have to deal with for both PoW and this new proposed approach.

Sure, but that's my point, I don't think the economic incentives for proof of storage are any better.

How can they not be better if they remove the need to burn a bunch of power 24/7 and produce waste heat? It's replacing an ongoing cost with a fixed one. In both cases you need a hardware investment (PoW mining hardware for one, persistent storage for the other) but the proof of work system ALSO has a large ongoing energy cost in addition to hardware.

GPUs or ASICs for hashing are going to have a limited life span just like your storage hardware, too.

In this case the storage hardware is more directly linked to the output proof. Much more hardware manufacturing would be involved.

In PoW you can lower your environmental impact by using clean electricity. For Proof of storage it would be 100% manufacturing.

I was looking into their whitepaper trying to find a counter argument to this. My conclusion is basically this:

The probability of being awarded a transaction is proportional to your reserved space 'plot' compared to the globally reserved space. Assuming it takes off, and enough 'overprovisioned' storage has been added to the pool, then it makes little sense to buy hardware just to farm storage-coins. The tiny increase in probability just wouldn't pay off.

This is the same proposition with PoW and hashes though, surely?

The main cost of compute is electricity.

The main cost of hard drives is intellectual property (ie. Paying for the R&D, patents and licenses for all the tech inside).

It isn't clear that "wasting" intellectual property is in any way bad for the wider economy. In fact, it probably just subsidies storage for the rest of us in the long term.

This doesn't sound right, it's very doubtful that all storage tech has royalty (unit-based) license terms.

Instead of electricity consumption the limiting factor will be raw materials to manufacture storage. Likely even worse for the environment given the energy and raw material cost to manufacture storage devices.

I was more thinking of the waste involved in manufacturing. If there's suddenly a much more direct profit motive to buying drives it's going to push more manufacturing, which is not the same as ongoing burning of energy for PoW, sure, but it's not going to be without side effects.

overall it would be significantly less, that is the point

I'm not convinced it would.

Just like PoW it makes sense to spend enough until a breakeven point. In thise case there is a direct incentive to buy as much storage as possible.