← OrreryThe Ecliptic is an Orrery publication

Cover Story · Chain Reaction desk

The Gas Limit Now Rises by Default

In 2025 Ethereum's gas limit moved only after about half of validators chose to raise it. Glamsterdam's gas-limit schedule makes the scheduled value every conforming client's default, so a majority now has to act to hold it back. On Sepolia, 39% of sampled proposers voted the limit down and it still sat at 99.9% of its 200M target.

~10 months
for mainnet's limit to climb from 36M to 60M in 2025, each step waiting until about half of validators had signalled for it
99.9%
of the 200M schedule delivered on Sepolia on 22 October (mean 199.82M across 47 blocks) although 18 of 46 sampled proposers voted the limit down
4.1 h
to ramp from 60M to 200M if every proposer follows the schedule; about 19 hours at Sepolia's 39% dissent, and never at 50% or more (Chain Reaction estimates)
242%
today's state creation as a share of the repriced state lane's target if mainnet stays at 60M and demand does not respond to price (Chain Reaction estimate)

The burden of action

The limit is still a vote. The default it starts from has changed.

010203040hours from 60M to 200M0% — unanimous4.1 h8% — mainnet sample4.9 h25%8.2 h39% — Sepolia sample18.9 h45%41.1 h50% or morenever — the limit falls back
Estimated hours for the gas limit to rise from 60M to 200M by the share of proposers voting it down, from SAT-0036's ramp table. Navy rows are the unanimous case and the observed samples; teal rows are reference points. The 8% and 39% rows are the down-step shares in that report's mainnet and Sepolia samples of 22 October; the mainnet row applies today's share to a hypothetical 200M schedule. All values are Chain Reaction estimates from the 1/1024 rule and 12-second slots; not illustrative.

Each block may move the gas limit by at most 1/1024 of its parent, so the limit follows whichever side holds most of the block proposals. Until now a higher limit needed a majority of validators to opt in. EIP-8261 makes a dated schedule the value every conforming client uses when its operator sets nothing, which hands the up-voters every operator who accepts defaults.

  • 2025 signalling for a higher limit: 10% on 19 December 2024, above 30% on 16 January, 52% on 4 February, when the limit left 30M for 36M.
  • The 45M step arrived on 22 July 2025 with nearly half signalling; 60M came at the end of November 2025, alongside EIP-7935's change of client defaults.
  • The schedule is advisory: clients log a warning when an operator sets more, and the 1/1024 rule remains the only validity rule.
  • From 60M, 200M takes about 1,234 consecutive full upward steps; with a share *d* voting down the time is about 4.1 ÷ (1 − 2*d*) hours (Chain Reaction estimates).

The evidence

A 39% dissent cost the schedule 0.09% of its value.

Glamsterdam activated on Sepolia on 6 October with a 200M schedule. In 47 consecutive blocks on 22 October every downward move was the largest the rule allows, yet the limit never fell below 199.22M. The down-voters appear as a sawtooth, not as a lower level.

  • Sepolia, 47 blocks: limit 199.22M–200.00M, exactly 200M in 11; transitions lowered / raised / held 18 / 23 / 5 of 46.
  • Sepolia blocks ran 27–46% full at a 15 wei base fee, so the network shows that clients accept 200M, not which resource fills it.
  • Mainnet the same afternoon: 60M in 19 of 25 blocks, about 52% full. Hoodi, four days before its 26 October fork, sat at 60M with blocks averaging about 54% full.
  • Two client releases were warned to default to 60M after Sepolia's fork; proposers still targeting 60M would produce exactly the observed pattern. The report calls this consistent with the data, not proven.

The decision left open

The mainnet schedule's first value sizes the new state lane.

0%50%100%150%200%250%242%60M194%75M161%90M145%100M97%150M73%200Mlane targetrealised gas limit
Today's state creation (326 MiB a day at a 60M limit, from EIP-8037) as a share of the repriced state lane's target at each realised gas limit, assuming demand does not respond to the higher price; coral bars are over target. Values are SAT-0036's own table (Chain Reaction estimates); in its elasticity cases the lane is under target from 100M upward at 0.3 (79–90%) and at every limit at 0.6. Not illustrative.

The state repricing (EIP-8037) applies in full from the fork block, and the lane it governs is half the gas limit divided by 1,530 gas per byte. No mainnet schedule has been set. A conservative start, the cautious choice for execution, is the aggressive one for state: the two lanes share one limit.

  • Jump to 200M at the fork, as on Sepolia and Hoodi: the lane opens 21–73% full and the limit arrives within hours.
  • Start at 60M and step 15M every ~89 days, as the EIP's own illustration does: the limit passes 150M after about 1.5 years, with the lane over target below 150M if demand does not respond to price.
  • No schedule on mainnet: the limit stays at the 60M default and today's state creation is 2.4 times the lane's target at zero elasticity.
  • The indicator to watch changes: the share voting *below* the schedule, not support for "a higher limit" (Chain Reaction estimates throughout).
The Ecliptic, Issue No. 015, cover analysis, drawn from Chain Reaction SAT-0036 (October 22, 2026), which builds on SAT-0033. Signalling shares, client defaults, schedule terms and block data are as reported in that report's cited sources (EIP-8261, EIP-7935, EIP-8037, Ethereum Foundation announcements, Etherscan samples of 22 October 2026); ramp times, sample averages and state-lane shares are Chain Reaction estimates on the relations the report states. Neither chart is illustrative.
THE ECLIPTICan Orrery publication
Content is produced by AI systems with human editorial oversight.