What a Bet Actually Costs on Sui: Gas, Measured Across 600 Mainnet Rounds
"Gas fees" is the first objection anyone raises about an on-chain casino, and it is a fair one. If every bet carries a network fee, then a game with a 3% house edge and a 3% fee is not a casino, it is a toll booth.
So we measured it. We pulled the gas summary from 600 successful bets settled on Sui mainnet — 300 coinflip rounds and 300 keno rounds — and looked at what players actually paid. The headline: the median bet cost 0.0040 SUI in coinflip and 0.0045 SUI in keno. At the SUI price when we ran this, that is about a third of a US cent.
The more interesting finding is what that number is made of, because almost none of it is what people mean by "gas".
What 600 mainnet bets actually cost
| Coinflip | Keno | |
|---|---|---|
| Median net gas | 0.003952 SUI | 0.004468 SUI |
| 10th percentile | 0.002554 SUI | 0.002711 SUI |
| 90th percentile | 0.008215 SUI | 0.008489 SUI |
| Cheapest observed | -0.073183 SUI | -0.122782 SUI |
| Most expensive observed | 0.027440 SUI | 0.013370 SUI |
Eight out of ten bets landed between 0.0026 and 0.0085 SUI. The spread is real but small in absolute terms, and it is not caused by network congestion in the way an Ethereum user would assume. It is caused by how many objects the transaction created and destroyed, which we will come back to.
And yes, some bets are free
The "cheapest observed" row is not a typo. Some bets have negative net gas: the player finished the transaction with more SUI than the fee took. The most extreme keno round in our sample returned 0.1228 SUI to the player in net storage rebate.
That is not a promotion or a rebate program. It is how Sui storage accounting works, and it is worth understanding.
Computation is a rounding error. Storage is the whole story.
Sui splits a transaction fee into two parts: a computation cost for the work the validators do, and a storage cost for the bytes the transaction leaves on-chain. When a transaction deletes an object, the storage deposit that object was holding is refunded as a storage rebate.
Here is how our 600 bets split across those three lines:
| Component | Coinflip (avg) | Keno (avg) |
|---|---|---|
| Computation cost | 0.000104 SUI | 0.000234 SUI |
| Storage cost | 0.058630 SUI | 0.059449 SUI |
| Storage rebate | -0.053820 SUI | -0.054807 SUI |
| Net paid | ~0.0048 SUI | ~0.0049 SUI |
Read that first row again. Running the actual game — the randomness draw, the payout logic, the settlement — costs about 0.0001 SUI in coinflip. That is a hundredth of a US cent. The computation genuinely is close to free.
What the player pays for is a temporary deposit on the objects a bet creates while it is in flight, most of which comes straight back when the round settles and those objects are cleaned up. The net figure is the difference between a large number and a slightly smaller large number, which is exactly why it moves around between 0.0026 and 0.0085 SUI, and why it occasionally goes negative when a round happens to clean up more state than it created.
Putting it next to the house edge
A fee only matters relative to what you are betting. Here is the median bet cost expressed as a share of stake, at a few common bet sizes, using the 0.0045 SUI median.
| Stake | Gas as share of stake |
|---|---|
| 1 SUI | 0.45% |
| 5 SUI | 0.09% |
| 10 SUI | 0.045% |
| 50 SUI | 0.009% |
| 100 SUI | 0.0045% |
At a 1 SUI stake, gas is 0.45% — noticeable next to a house edge of a few percent, though still far smaller than it. By 10 SUI it is under a twentieth of a percent, and past that it stops being a line item at all. Gas on Sui is a fixed cost per bet, so it hurts small stakes and disappears on large ones. That is the opposite of a percentage-based deposit fee or a spread, which scale with you forever.
It is also worth being precise about what this fee buys, because it is not the same thing an off-chain casino charges you for. It pays for the settlement itself: the draw, the payout, and a permanent public record that the round happened and how.
Why this is a Sui number, not a blockchain number
None of this generalises to "on-chain gambling" as a category. It is specific to how Sui prices gas and how its storage fund works. Two design choices do most of the work.
- Computation is priced per unit of work, not auctioned. There is no fee market bidding war between users, so a bet does not get more expensive because someone else is minting an NFT collection in the same block.
- Storage deposits are refundable. Short-lived objects — exactly what a bet is — cost almost nothing over their lifetime, because the deposit comes back when the round settles.
Add native on-chain randomness, which means the draw does not need an external oracle call or a second transaction to resolve, and a full bet round fits inside one cheap transaction. We wrote up why we chose Sui for iGaming in more detail if you want the architectural version of this argument.
How to check this yourself
Every figure here comes from public data. Each bet is a mainnet transaction with a digest, and the gas summary — computation cost, storage cost, storage rebate — is part of its effects. Any Sui explorer will show you all three for any round you have played, and our walkthrough on verifying a bet on-chain covers reading the rest of the transaction.
If you want to see how the fee sits against the edge on a specific game, we broke down house edge and RTP for on-chain casinos separately. Every live game is at suigar.com/games.
Method
We queried the Sui mainnet GraphQL endpoint for the 300 most recent successful transactions calling the Suigar coinflip module and the 300 most recent calling the keno module, then read gasSummary from each transaction effects. Net cost is computation cost plus storage cost minus storage rebate, converted from MIST at 1 SUI = 1,000,000,000 MIST. Percentiles are over successful transactions only. The USD conversion in the opening paragraph uses a SUI spot price of $0.787 from the Coinbase public price API on 5 September 2026; SUI is volatile, so treat that figure as an illustration and the SUI-denominated numbers as the measurement.






