Ethereum

In brief

Ethereum is a programmable blockchain first proposed by Vitalik Buterin in a draft circulated in late 2013, with the canonical whitepaper published in 2014, and launched as the Frontier network on 30 July 2015. Its native asset, ether (ETH), pays for computation. Accounts are either key-controlled (externally owned) or code-controlled (contracts). Transactions invoke the Ethereum Virtual Machine, which meters work in gas. Tokens on Ethereum are usually separate contracts following interfaces such as ERC-20; that is not how tokens work on every network. In September 2022, The Merge switched block production from proof-of-work mining to proof-of-stake validators. It did not reset balances, rewrite history or replace the EVM. Scaling today is largely rollup-centric. Protocol changes go through public Ethereum Improvement Proposals and multiple independent clients. Saying “a company controls Ethereum” collapses that process into a corporate product, which it is not.

Why Ethereum was proposed

Bitcoin showed that a network could maintain a ledger without a mint. By 2013, people were already grafting extra meaning onto Bitcoin with colored coins, name systems and limited scripts. Buterin’s whitepaper argued that Bitcoin Script was too constrained for the applications people wanted: it lacked loops, fine-grained value control, persistent state and easy access to blockchain data.

Ethereum’s proposal was a chain whose native programming language was general enough to encode “arbitrary state transition functions.” Instead of a new blockchain or a clumsy meta-protocol for each idea, developers would write a contract in a few lines and let the same set of nodes execute it. The paper calls this a next-generation smart-contract and decentralized-application platform.

That is a design for one network. It is not a definition of “Web3,” and later systems chose other account models, virtual machines and token architectures.

People, paper and launch

Vitalik Buterin wrote the introductory paper. Official ethereum.org pages preserve two different milestones: the history timeline labels an introductory paper “released” on 27 November 2013, while the current whitepaper page calls the canonical paper published in 2014. The most precise shorthand is therefore late-2013 proposal/draft, 2014 canonical whitepaper, not one uncontested publication date. The site also notes that the living protocol has moved far beyond that text. Buterin is often called Ethereum’s inventor. He is not a chief executive who operates the chain.

Gavin Wood wrote the Yellow Paper, a formal protocol specification, dated April 2014 on the same timeline. An ether sale ran for 42 days from 22 July 2014. Other early contributors included people later associated with clients, companies and the Ethereum Foundation. Treating any one of them as “the owner” misreads how the network ships.

Frontier, the first live mainnet release, is dated 30 July 2015 on ethereum.org’s fork timeline. An Ethereum Foundation blog post days earlier described Frontier as a bare-bones implementation for technical users, emerging from consensus rather than a central launch button, with a temporary low gas limit to thaw the chain. Homestead followed in March 2016. Later named forks changed fees, economics and, eventually, consensus.

Programmable blockchain, accounts and transactions

Ethereum’s state is a set of accounts, each with a 20-byte address. The traditional documentation distinguishes two types:

• Externally owned accounts (EOAs) are key-controlled accounts. Creating one is free. They can start transactions.
• Contract accounts are controlled by code. Creating one costs gas because it occupies storage. They run when they receive a call. They do not have a private key.

That split is still the right beginner picture, but it is no longer an absolute invariant. EIP-7702, activated in the Pectra upgrade on 7 May 2025, lets an EOA install a code-delegation pointer to already deployed code. Contrary to an early version of the proposal, the final delegation persists after the transaction until the key holder replaces or clears it. The account remains able to originate key-authorized transactions. The old “EOA never has code behavior” rule is therefore too sharp in practice. This article does not treat that feature as the way every blockchain account works.

Every account stores a nonce, an ETH balance (in wei; 10^18 wei = 1 ETH), a code hash and a storage root. Users hold keys, not the ether itself; balances live in the state trie.

A transaction is a signed instruction from an EOA. It specifies a recipient, value, data, and gas parameters. Contract-creation transactions deploy bytecode. Message-call transactions transfer ETH and/or run code. Contract-to-contract calls are internal messages, not separately signed user transactions.

This account model is not universal. Bitcoin tracks unspent outputs. Solana uses program-owned data accounts and, for tokens, mint and token accounts under a Token Program.

The EVM and smart contracts

The Ethereum Virtual Machine is the shared execution environment. Every full node runs the same bytecode so that applying a block produces the same new state. ethereum.org describes Ethereum as a distributed state machine: Y(S, T) = S' — a valid prior state plus valid transactions yield a new state.

The EVM is a stack machine with 256-bit words. It has transient memory, optional transient storage for the duration of a transaction, and persistent contract storage. Operations cost gas. If a call runs out of gas, its state changes revert; the fee for work already attempted is still paid.

A smart contract is that bytecode plus storage, addressed on-chain. It is not a legal contract and not a background server. It runs only when a transaction (or another contract) calls it, and every validating node re-executes that call. The original whitepaper compared contracts to autonomous agents that act when poked.

Users typically write Solidity or another high-level language; compilers emit EVM opcodes. Multiple execution clients implement the EVM so that no single program defines the chain.

Gas and ETH

Gas measures computational effort. The fee is gas used times the price per unit. Fees are paid in ETH, usually quoted in gwei (10^9 wei).

Since the London upgrade (August 2021, EIP-1559), a protocol base fee is burned and a priority fee (tip) goes to the block producer. Users may set a maxFeePerGas. 21,000 gas is the classic intrinsic cost of a simple ETH transfer from an ordinary key-controlled account. Contract calls need more, and newer account behavior or transaction types can change the picture. The gas limit is the user’s ceiling for that transaction. Too low, and the call fails; unused gas in a successful call is not all consumed.

ETH is the native asset. It is not an ERC-20 token. It pays gas, can be sent directly between accounts, and after The Merge is also the asset validators stake. Issuance and burning both affect supply; this article does not treat that as an investment thesis.

Tokens: ERC-20 is an Ethereum interface, not a universal model

ERC-20 (2015, Fabian Vogelsteller and Vitalik Buterin) defines a common contract API: totalSupply, balanceOf, transfer, transferFrom, approve, allowance, plus Transfer and Approval events. A token is a contract that keeps a ledger in storage. Wallets and exchanges integrate one interface instead of a new one per asset.

That design is specific to Ethereum-style contract platforms.

• On Solana, fungible and non-fungible assets are mint accounts and token accounts owned by a Token Program (or Token-2022). Balances are not, by default, mappings inside a custom per-token contract.
• Bitcoin has no ERC-20 equivalent as a native standard; extra asset schemes are overlays or later protocol experiments.
• Other chains mint assets at the ledger layer rather than as user-deployed contracts.

Calling every crypto asset “an ERC-20” erases those differences. Even on Ethereum, NFT-style assets use other interfaces (for example ERC-721), and ETH itself is not an ERC-20.

Proof of stake, validators and the two layers

Ethereum launched with proof of work. On 15 September 2022 it switched to proof of stake.

A validator still needs a 32 ETH activation deposit (MIN_ACTIVATION_BALANCE). EIP-7251, included in Pectra, raised the maximum effective balance a single validator can have to 2,048 ETH so operators can consolidate stake. That is a ceiling, not a requirement: nobody must deposit 2,048 ETH to validate. The validator also runs three pieces of software: an execution client, a consensus client and a validator client. Validators check blocks, attest, and sometimes propose. Dishonest behavior can be slashed.

Time is divided into 12-second slots and 32-slot epochs. A proposer is selected per slot; committees attest. Finality uses checkpoint votes: when votes representing at least two-thirds of active stake link checkpoints, an epoch can become justified and then finalized. Reverting finalized history would require at least one-third of the total stake to become slashable.

After The Merge, a full node is two cooperating programs:

• the execution layer (the former “Eth1” client) handles transactions, the EVM and state;
• the consensus layer (the Beacon Chain) handles proof-of-stake, attestations and fork choice.

They talk over an authenticated Engine API. Running a non-validating node does not require 32 ETH.

The Merge: what changed and what did not

ethereum.org dates The Merge to 15 September 2022 (Paris upgrade, block 15,537,394). The Beacon Chain had already been live since 1 December 2020, reaching consensus on validators in parallel with proof-of-work mainnet. The Merge joined them and turned off mining.

Changed: block production method; energy use; issuance rules for new ETH; the requirement to run both client types; slot timing (12 seconds). ethereum.org gives an estimated energy-consumption reduction of about 99.95%. A separate CCRI report commissioned by ConsenSys, using literature for proof of work and hardware/client measurements for proof of stake, estimated a reduction in annualized electricity consumption of more than 99.988%. Both are estimates, not direct meter readings of every node or former miner.

Did not change: account balances; contract code; ETH as the native asset; the EVM; the need to pay gas; historical data since 2015. There is no separate “ETH2” token. Users did not migrate funds.

The Merge did not lower gas fees or finish scaling. It also did not, by itself, enable staking withdrawals; those arrived later in the Shanghai/Capella (“Shapella”) upgrade on 12 April 2023.

Scaling and rollups

On-chain demand raises gas prices. Ethereum’s current scaling path, described in ethereum.org’s scaling docs, is mostly off-chain execution with on-chain security, not a return to the old plan of many execution shards.

A rollup executes transactions off the main chain, then posts data (and, depending on the design, a proof) to Ethereum so that mainnet can enforce correctness.

• Optimistic rollups assume batches are valid and rely on a challenge window and fraud proofs.
• Zero-knowledge rollups post a validity proof with the batch.

Sidechains, plasma and validiums exist too; they do not all inherit Ethereum security the same way. Layer 2 is, in ethereum.org’s narrower usage, an off-chain system that ultimately depends on layer-1 consensus.

This architecture is an Ethereum roadmap choice. It is not how every blockchain scales.

Governance and why “a company runs Ethereum” is wrong

No single process both writes the code and forces nodes to run it.

• Ethereum Improvement Proposals (EIPs) specify proposed changes. Anyone can write one, but a proposal is not automatically adopted: core changes require public discussion, implementation across clients and broad enough support for operators to coordinate on the upgrade.
• Client teams (multiple execution clients, multiple consensus clients) implement specs independently. Client diversity is treated as a safety property.
• Node operators and validators choose which software to run. An upgrade that they refuse does not become the chain they follow.
• The Ethereum Foundation is a non-profit that funds research, education and ecosystem work. Its own site says it supports the ecosystem “without controlling it,” and that it is not a typical tech company. ethereum.org states there is no company called Ethereum that manages accounts or holds user funds.

Influential researchers and funders matter. They are not a kill switch. A useful comparison is the internet: vendors, standards groups and operators matter; none of them is “the company that is the internet.”

Forks make the point concrete. In July 2016, after a major application (The DAO) was exploited, the community split over whether to change state to return funds. The fork that moved those funds is today’s Ethereum; the chain that refused became Ethereum Classic. That episode is a reminder that “immutability” includes social choice about which rules to keep, not only hash links.

Historical markers worth knowing

Only a few events change how a beginner should picture the system:

• 2013–2014: whitepaper, Yellow Paper, ether sale.
• 30 July 2015: Frontier mainnet.
• 20 July 2016: DAO fork and the Ethereum / Ethereum Classic split.
• 1 December 2020: Beacon Chain genesis.
• 5 August 2021: London / EIP-1559 fee market.
• 15 September 2022: The Merge.
• 12 April 2023: Shapella withdrawals.
• 2024–2025: Dencun (cheaper rollup data), Pectra (including EIP-7702 account features).

Most other named forks adjusted gas, mining rewards or the old proof-of-work “difficulty bomb.” They matter to operators more than to a conceptual overview.