# Ethereum studies cutting finality time to four minutes

> An EIP-8453 draft separates head votes from finality votes so that Ethereum goes from 16 to 4 minutes of waiting per transaction.
> By José Maldonado · Ethereum · 2026-10-25

A transaction on Ethereum can enter a block in twelve seconds and take more than a quarter of an hour to be closed forever. Five protocol researchers registered on October 21 a draft dubbed Decoupled Consensus, which already carries the number [EIP-8453](https://github.com/ethereum/EIPs/pull/12566) and which seeks to reduce that second wait to a quarter, from about 16 minutes to about 4.

The draft is signed by Ben Edgington, who coordinates fast finality research at the Ethereum Foundation, along with four other signatories. It does not touch block size or the price of gas. It separates the two tasks that today share a single validator vote: choosing which will be the next block and sealing old blocks so that no one can rewrite them.

The difference between 16 minutes and four matters in services that move money between networks. A bridge or an exchange advances funds to the user and waits for Ethereum to close the account; every minute of that wait is immobilized capital and a fee that someone ends up paying. Those who operate these services decide daily how long they leave an operation in limbo before considering the money good. The [live measurement of Ethereum's finality](https://openchainbench.com/benchmarks/l1-finality/ethereum) is currently around 15.9 minutes median, with 12.8 as the ideal case, so the four-minute target is not a cosmetic adjustment.

## Speeding up Ethereum's finalization

The current mechanism combines two pieces. Casper FFG seals the chain's checkpoints and LMD-GHOST chooses the head of the fork, and both feed on the same vote that each validator casts per epoch. Since that vote serves two jobs at once, sealing cannot go faster than the rotation that produces it.

The draft breaks the knot in two paths. The available chain uses a committee of about 512 validators that casts head votes to choose the next block and mark a confirmed tip. Finality goes on another track, with attestations from the full validator set that advance upon reaching two-thirds of active stake, without waiting for the epoch to close.

The text also changes the reinforcement that each block's proposer receives today and admits a new risk: by reducing the committee that votes the head, that group becomes an easier target for a denial-of-service attack.

Expiring locks are the other piece. A block becomes firmly locked when at least half of the full committee votes in its favor on time and without double messages, and that mark protects the confirmed tip if participation falls. Contradictory attestations can be penalized, although the exact conditions and penalties remain unspecified.

The text itself admits what is missing. The timing parameters, the rewards and punishments and parts of the security analysis appear marked as pending, and the fourfold reduction is conditional on the choice of those values.

Edgington wrote in the draft's log that many parts remain undecided and that the immediate goal was to obtain an EIP number to enter the protocol's formal improvement circuit.

## Confirmed is not final, and that difference costs money

An Ethereum block is proposed every 12 seconds and anyone looking at the chain sees it almost instantly. That is confirmation, not finality. Finality arrives when validators commit not to rewrite that block, and in between there is room for a reorganization that returns a transaction to the queue or erases it from the history everyone takes as good.

Transfer providers build their business on that gap. [Circle waits for two confirmations and about 20 seconds](https://developers.circle.com/cctp/concepts/finality-and-block-confirmations) to attest a fast USDC transfer from Ethereum, while the standard route awaits hard finality and takes minutes.

The first charges a fee and the second does not. On Ethereum, the fast route costs [one basis point, 0.01%](https://developers.circle.com/cctp/concepts/fees), which is 10 dollars for every 100,000 transferred before counting network gas.

The limit is also a business one. Circle caps the total value of fast transfers awaiting settlement, and each closed operation frees room for the next ones. With finality at four minutes, that cap recycles more times per hour and the same credit line supports more volume without the company having to expand it.

Operation status

What it guarantees

Typical time on Ethereum

Included in a block

That a node sees it; it can still be reorganized

12 seconds per slot

Attested on fast route

That Circle signs the transfer with two confirmations

About 20 seconds

Final

That validators do not rewrite it

15.9 minutes median

## The idle money of bridges and exchanges

The arithmetic behind the draft is treasury arithmetic. A service that advances a million dollars per hour has 16,667 dollars at stake per minute; with 16 minutes of waiting, about 267,000 dollars remain pending collection, and with four minutes, about 67,000. That money does not disappear from the balance sheet, but it stops being tied up twelve minutes longer.

For an exchange there are two ways to decide and neither is free. If it credits the balance to the client after finality, the client waits; if it credits it before, the house assumes the risk of a reorganization and covers it with its own limits and buffers. Shortening the wait reduces that buffer without changing the rule of when an operation is settled.

Rollups play another league. The standard withdrawal of an optimistic rollup waits for its challenge window, which is measured in days, so the finality cut does not move it by a single minute.

Where it does show is in the cost of publishing data on layer one, which this year has given scares. [On October 4 the price of blob gas rose by around 400% and rollups paid the bill](https://observatorioblockchain.com/en/ethereum/ethereum-blob-fees-400-l2-limites/), an expense that ends up with the user of the layer-two network.

## Glamsterdam is delayed on another track

Finality does not travel in Ethereum's next upgrade. Glamsterdam had a date on Hoodi for October 26 at 17:42:48 UTC, in epoch 132,352, and that appointment no longer exists: [the Ethereum Foundation postponed the activation](https://blog.ethereum.org/2026/10/22/glamsterdam-hoodi-delay) after the ACDE #248 developers' call, with no replacement date.

The reason is gas, not consensus. Developers want to investigate how EIP-8037, the new state-creation gas calculation, interacts with contracts already deployed on the mainnet. The postponement also withdraws the schedule that was going to raise Hoodi's gas limit to 200 million in that same epoch.

The two efforts go along different paths and neither waits for the other. Glamsterdam touches the price of state and the division of tasks between whoever proposes a block and whoever builds it; EIP-8453 changes the way of voting and has no activation date on any network. One slowing down does not speed up or delay the other, but both depend on the same fork bureaucracy and on clients implementing what is signed.

The Sepolia trial serves as a reference. [The 200 million gas blocks were tested there on October 6](https://observatorioblockchain.com/en/ethereum/glamsterdam-sepolia-200-millones-gas-prueba/) and moved between 52 and 92 million, from 26% to 46% of capacity. Hoodi was going to repeat the test on October 26 and now has no time.

## A change for those who move money

None of this changes tomorrow. The text remains in Draft status, needs a consensus fork, implementations in the clients and parameters that today are blank, and not even the fourfold reduction is guaranteed, because the draft itself conditions it on the analysis of those values. The first realistic milestone is seeing it on a test network with a name and epoch.

Meanwhile, the data worth watching is not the promise of four minutes but the gap between confirmation and finality in the services one uses. An exchange that credits in 20 seconds assumes the risk that slow finality transfers to it, and a bridge that charges for advancing funds puts a price on those 16 minutes. When finality drops, the question will be whether that price drops with it or whether the margin stays where it is.
