Empirical Study: bank/multi-send Gas Surcharge in Gaia v27.0.0

TL;DR

We ran 5 real transactions on the provider testnet to empirically
verify the multi-send gas surcharge introduced in Gaia v27.0.0
(PR #3961). We confirmed the quadratic curve — and found an additional
linear component of ~8,628 gas per recipient from SDK store I/O.

Complete empirical formula: gas_total ≈ 95,302 + 8,628 × n + 300 × n²


Background

Gaia v27.0.0 introduced a quadratic gas surcharge on MsgMultiSend
to mitigate spam:

gas_surcharge = 300 × n²

We wanted to verify this in practice, so during Testnet Tuesday
(March 2026) we ran transactions with 1, 10, 20, 49, and 100
recipients on the provider testnet, recording the actual gas_used
from the public REST API.


Results

n gas_used (real) Theoretical 300n² Ratio
1 (base) 103,963
10 211,587 30,000 3.59x
20 387,909 120,000 2.37x
49 1,238,130 720,300 1.57x
100 3,958,152 3,000,000 1.28x

The ratio converges toward 1x as n grows, confirming the quadratic
term dominates at scale.


Key Finding

Beyond the documented 300 × n² surcharge, there is a consistent
linear component of ~8,628 gas per recipient representing the
Cosmos SDK store I/O cost (read account + validate balance + write
state). This slope is constant within ±0.2% across all measurements.

The complete empirical formula:

gas_total ≈ 95,302 + 8,628 × n + 300 × n²

Prediction error < 0.1% on all measured points.

Inflection point at n=29: below this threshold, the linear SDK
cost exceeds the quadratic surcharge. Above n=29, the quadratic
term dominates.


On Deterrence

Other validators also ran independent tests during Testnet Tuesday
and raised a valid point: at current ATOM prices, even 400 recipients
costs ~0.26–0.32 ATOM, which is cheap in absolute terms.

Our view: the mechanism is dynamic pricing, not a firewall. At
ATOM=$100, that same transaction costs ~$25. Scaling spam to thousands
of transactions becomes economically irrational. Near ~490 recipients,
the block gas limit (75M) rejects the transaction outright regardless
of fee — a hard ceiling.

For legitimate use cases (50–100 recipients), costs remain very
reasonable at any realistic ATOM price.


Verified Transactions

n TX Hash
1 60886AFCE17610B5EB1E0E8E33DA84C2B7D7EBD8A7541CC5F6CAE7965A6CA873
10 5B704EBDD7739698BFB1F67F6CFFB12CE148B208072CD1030A988B8CFECE55E7
20 5CD0F241C0FF48C7D26D1AF90FCAD6F3BE9A721B592B4D5BF828E19E568F79BC
49 499B5D156CC7447E625932A0EF63C78E49C94D7376D1D280048E82D66E8967E9
100 A19877023233F99782AA384355649EC2F39F4F6D091163DDE0110AF8EB68FB79

Explorer: mintscan.io/ics-testnet-provider/tx/<HASH>

Full study with charts:
github.com/Cumulo-pro/cumulo-cosmoshub-infra/blob/main/studies/gas-surcharge-multisend-2026-03/README.md


Cumulo · Cosmos Hub Validator · Testnet Tuesday March 2026

1 Like