Ethereum’s Mid-Air Engine Swap: The Merge and the Rules It Changed

A new engine alongside the old one

In January 2014, Vitalik Buterin published Slasher, an early proposal for penalizing conflicting participation in proof-of-stake. Its opening was cautious: he was exploring a possible tool, not announcing that Ethereum had selected it as its mining algorithm. That distinction makes the eventual Merge more interesting. The distance from a research idea to a working replacement for a live network was filled with design decisions that the first proposal had not settled.^[5]

One problem was how to make contradictory votes costly. Buterin’s 2017 proof-of-stake FAQ examined the incentive to support competing histories when doing so requires little extra computation. Ethereum’s later Beacon Chain combined Casper FFG finality with a fork-choice rule, LMD-GHOST. Its design distinguished ordinary performance penalties from slashing for specified conflicting messages. Slashing losses could grow with correlated misconduct; the rules were more discriminating than an automatic confiscation of every offending validator’s entire balance.^[3][11]

The Beacon Chain launched on December 1, 2020. The announced genesis condition required at least 16,384 deposits of 32 ETH, a minimum of 524,288 ETH. Those were validator deposits, not a census of individual people or the exact final deposit total. Initially, the chain tracked its own validators and balances while Ethereum mainnet continued handling applications under proof-of-work. Running the systems beside each other gave the replacement consensus a separate operational life before it inherited mainnet’s transactions.^[1][6]

For the people supplying that stake, the separation had a concrete consequence: deposits could enter the new system, but withdrawals were not part of its initial operation. Nor would The Merge itself unlock them. The eventual handover therefore changed responsibilities before it completed the withdrawal path. This was an upgrade with deliberately limited scope, built around moving consensus first rather than releasing every planned feature together.^[11][4]

The threshold that handed over the chain

The switch needed a rule that clients could apply to the same history. EIP-3675 explains why a preset block number was unsafe for this particular transition: an attacker could build a minority-work fork that reached the chosen height and try to place the first proof-of-stake block on it. The specification instead used accumulated proof-of-work difficulty. Its purpose was to constrain which history could qualify for the handover, not merely to put a dramatic number on a countdown screen.^[2]

Ethereum mainnet’s Terminal Total Difficulty was 58750000000000000000000. A terminal proof-of-work block had to reach or exceed that threshold while its parent remained below it. The upgraded rules rejected subsequent proof-of-work blocks beyond that terminal block. From there, proof-of-stake fork choice took over. These were explicit validation conditions, including provisions for competing terminal blocks before finalization, rather than a promise that every machine would react at exactly the same instant.^[2]

The August 24, 2022 mainnet announcement described two stages: Bellatrix on the consensus layer, then Paris on the execution layer at the difficulty threshold. Operators had to update both clients and connect them through the Engine API. The announcement reported completed public-testnet upgrades, but also listed client releases and operational warnings. Preparation meant making different implementations work together; the existing contracts and their users were the reason that coordination mattered.^[4]

This EIP deprecates Proof-of-Work (PoW) and supersedes it with the new Proof-of-Stake consensus mechanism (PoS) driven by the beacon chain.^[2]

— EIP-3675

A copied ledger and the assets behind it

The financial applications above those clients faced another question: what would their assets mean on a fork? On August 9, 2022, Circle said it intended to support only Ethereum’s proof-of-stake chain after The Merge. Its USDC announcement treated a single supported version as essential to continuity. A copied token balance therefore did not automatically carry the issuer’s support with it. This was a decision about USDC, not a command that could stop other software from producing blocks.^[7]

Tether published its own position that same day. It intended to support tether tokens on the post-Merge proof-of-stake chain and cease support for the proof-of-work chain. The company framed this around avoiding disruption and confusion for users and DeFi platforms. For an application relying on those tokens, issuer policy was part of the asset’s practical meaning, even though the balance itself appeared in an open ledger.^[8]

EthereumPoW nevertheless existed as a separate fork. Circle’s November 2023 explanation of its USDC contract upgrade explicitly referred to the creation and operation of that chain while reiterating which Ethereum assets it supported for redemption. The contrast is revealing: a network can continue producing its own history without inheriting every service or financial relationship attached to the chain it copied. Survival as software and continuity of an issuer-backed asset are different questions.^[10]

This is where the familiar engine-swap metaphor meets an economy. Consensus software decides which blocks its participants accept; an issuer decides which version of its liability it supports. The Merge brought those choices into the same historical moment without making them identical. Its story cannot be reduced to a contest between computers and applications: it exposed how a ledger’s rules and the institutions using that ledger depend on different forms of coordination.^[7][8][10]

What changed when mining ended

Ethereum completed The Merge on September 15, 2022. Block production and finalization were distinct steps: the advance announcement expected the first post-threshold block to finalize about two epochs, roughly thirteen minutes, later under normal network conditions. The achievement was continuity of the existing chain under a replacement consensus mechanism. Users retained their history and balances; the network did not require them to exchange ordinary ETH for a newly issued replacement coin.^[1][4]

The energy result was substantial. Ethereum.org puts the reduction in Ethereum’s energy consumption at approximately 99.95%, because competitive proof-of-work mining was no longer required. That estimate describes this network’s energy use. Carbon emissions also depend on how electricity is generated, and a worldwide percentage would need its own measurement basis. The enduring engineering result is already clear without those extra comparisons: securing Ethereum no longer required the former mining competition.^[1]

The issuance account changed as well. Ethereum.org’s historical comparison estimates an approximately 88.7% reduction using its stated staking assumptions, close to the commonly rounded figure of 90%. Mining issuance ended while validator issuance continued. The base-fee burn introduced before The Merge remained a separate force. Whether ETH supply increased or decreased depended on the balance between issuance and burning, so a successful consensus upgrade did not guarantee permanently falling supply or a particular market price.^[9]

The Merge left further work ahead: it was not a capacity expansion designed to cut gas fees, and staking withdrawals arrived through a later upgrade. Its lasting accomplishment was narrower and demanding enough—moving an operating chain to a different security model while carrying its history forward. That model still depends on participation, functioning software and its stated security assumptions. Changing the engine was a milestone in maintaining a public network, not the end of maintaining it.^[1][11]