Ethereum Paymaster Gas: Who Pays?
The gas tank on your swap is empty and the transaction still lands — that is not magic, that is a paymaster fronting the ETH and taking on the settlement risk. If you think account abstraction solved the fee problem, you are half right. The paymaster solved the UX problem and inherited every accounting nightmare that comes with it. 💸
The Sponsor Is the Payer, Not the User
Strip away the bundler marketing and the mechanics are brutal: a smart account sends a UserOperation, the bundler simulates it, and the paymaster contract posts a deposit to the EntryPoint that covers the maximum possible gas cost. That deposit is the real money. The user pays nothing at execution time. The Ethereum Paymaster Gas bill lands on whoever funded the paymaster's stake — a dApp treasury, a game studio, a market maker, or a relayer running a sponsored tier. Everything else is a reimbursement agreement layered on top.
How the Prefund Math Actually Works
Before a single op executes, the EntryPoint validates the paymaster and pulls the max fee into escrow: gas limit multiplied by maxFeePerGas, plus the verification overhead. If the actual execution costs less, the surplus is refunded. If the op reverts, the paymaster still eats the consumed portion. That asymmetry is where sloppy sponsors bleed out — a popular sponsored faucet can be drained in hours by bots that submit reverting ops purely to burn sponsor liquidity. ⚠️
The Four Sponsorship Models and Their Risk Profiles
First, full sponsorship: the dApp absorbs 100 percent and users never touch ETH. Great for onboarding, catastrophic without rate limits. Second, conditional sponsorship: the paymaster covers only ops that pass an on-chain allowlist or a signed policy check. Third, token-based payment: the user pays in an ERC-20 and the paymaster swaps to cover ETH gas, exposing the sponsor to slippage and oracle lag. Fourth, deposit-and-refund schemes where the user pre-funds a credit balance and the paymaster draws it down. Each model shifts Ethereum Paymaster Gas liability to a different party, and each one is a distinct attack surface for griefing.
Where the Money Actually Leaks
Verification gas is the silent killer. Paymasters that perform signature checks, oracle reads, or allowlist lookups inside validatePaymasterUserOp pay for that compute on every single op — and they pay it before knowing whether the op will succeed. A paymaster doing three storage reads per validation at 21000 gas overhead per op is lighting real ETH on fire across thousands of transactions. The fix is aggressive off-chain signing: pre-approve ops with EIP-712 signatures so on-chain validation collapses to a single ecrecover. Every storage slot you remove from the validation path is a direct subsidy back to the sponsor.
Stake and Reputation Are Collateral, Not Credit
A paymaster must stake ETH in the EntryPoint to be accepted for ops that touch storage. That stake is not a payment — it is a bond the protocol can seize if the paymaster breaks rules or fails to cover post-op costs. Operators who confuse stake with budget get wiped when postOp reverts, because the EntryPoint still charges for the op and the paymaster's deposit absorbs the loss. Keep the entryPoint deposit topped up separately from the stake, and monitor balance drains on a block-by-block basis. 🧮
Throttling Is the Only Real Defense
No sponsor survives without per-account and per-IP gas ceilings enforced off-chain before the UserOperation is ever signed. Reputation scoring, cooldown windows, and dynamic per-op gas caps based on rolling 24-hour spend are the difference between a controlled subsidy and a public drain. The hard truth: whoever holds the paymaster deposit signs every op, and the chain does not care about your intent. It only cares about the balance at validation time.