Reference
Rent and cost
Soroban state is rented by size and by time. What follows is what we measured paying, not a formula quoted from elsewhere.
What you are paying for
Rent is a function of how large an entry is and how far into the future you are extending it. Extending the same entry twice as far costs roughly twice as much; extending an entry twice the size costs roughly twice as much. The fee is charged once, at submission.
There is no subscription and no ongoing charge. An entry with a large remaining TTL is not costing you anything per day — you have already paid for that time.
What we paid
| Extension | Fee charged | Note |
|---|---|---|
| +1,000 ledgers | 5,064 stroops | Expiry moved +1,002 including inclusion delay — the target is absolute, so ledgers that close while the transaction is in flight count. |
| engine extension | 44,725 stroops | An unattended run acting on a threshold it was configured with. |
Both are testnet, recorded with their transaction hashes on the extension history. They are receipts, not a price list: what an extension costs you depends on your entries.
Why batching is cheaper
Every submission pays a base fee regardless of how far it extends. Extending by thirty days once therefore costs less than extending by one day thirty times, and the difference is thirty base fees rather than anything about the rent itself.
This is why the default extension target is 518,400 ledgers, about thirty days, and why the act-now threshold is about one day: the gap between them is the window in which a single extension can be planned, simulated and submitted without hurry.
The estimate is an estimate
scan --cost prices an extension by simulating it against the network, so it reflects current pricing rather than a formula in this tool. Nothing is submitted.
Rent pricing varies with network state. We have measured it differ by about 18% between days, so treat a cost figure as a reading with a timestamp. If a budget depends on it, take the reading close to when you will submit.
Where the money actually goes
On guinea-pig A, one shared code entry is 8,116,648 of 8,264,289 stroops — 98% of the rent across all four of its entries. Scanning only the instance and code entries, that same entry is 99%: identical rent, different denominator.
The practical consequence is that optimising anything other than the code entry is usually rounding error, and that the entry worth watching is shared with every other contract built from the same Wasm. How state archival works
Next