Conventional storage is rented by the month and stops when payment stops. Decentralised storage networks generally cannot work that way, and the difference shapes everything about how they charge.
Nobody is in a position to send an invoice
A hosting company knows who its customer is and can suspend an account that stops paying. A permissionless network has no such relationship with the uploader.
The uploader may vanish immediately after posting the data. The nodes holding it still incur cost every month it remains.
So the network needs a way to fund storage from the moment of upload, without relying on anyone continuing to care about the file.
Endowment models pay from an upfront sum
One approach takes a single payment at upload and treats it as a principal that funds storage over a long horizon. Nodes are paid out of that pool over time.
The arithmetic depends on the cost of storage hardware falling faster than the pool is drawn down, which has historically held but is an assumption rather than a guarantee.
The user experience is unusual: one payment and no renewal. The cost is that the model is only as durable as the trend it depends on.
Contract models keep a recurring relationship
The alternative keeps storage as an ongoing agreement between the uploader and specific providers, renewed at intervals and priced by the market at each renewal.
Providers post collateral and lose it if they fail proofs, so the incentive to keep data intact is enforced financially rather than by reputation.
This tracks real costs more honestly, but it puts renewal on the user. Data disappears if nobody keeps the agreement alive.
Proving storage is the hard part
Either model needs a way to check that a provider still holds the data without transferring it. A challenge asks for a value derived from a random slice.
Answering requires possession, and the answer is small. Repeated challenges over time make continuous storage the cheapest strategy for the provider.
The proof system itself consumes resources, and its cost is folded into what users pay. Verification is not free, it is part of the price.
Retrieval is priced separately from storage
Holding data and serving it are distinct services with distinct costs. A node can hold a file cheaply and still have no reason to send it to whoever asks.
Networks therefore separate a retrieval market from a storage market, with bandwidth paid at the moment of the request.
Users comparing these networks to conventional hosting often miss this split, and find that the headline storage price describes only half of what they need.