The Merge: From PoW to PoS

On September 15, 2022, Ethereum accomplished what many thought impossible: it replaced the engine of a plane in mid-flight. A network worth $200 billion switched its consensus mechanism from proof-of-work to proof-of-stake with zero seconds of downtime. Energy consumption dropped 99.95% - from the level of an entire country to the level of a small office. How did engineers plan and execute the largest migration in blockchain history?

• **Energy:** The Merge cut Ethereum's energy consumption from ~112 TWh/year (comparable to the Netherlands) to ~0.01 TWh/year - the ESG critics' main argument against blockchain was eliminated in a single block
• **Staking economy:** after The Merge, more than 30 million ETH (~$100B+) is locked in staking, creating the largest DeFi market with 3–5% annual yield and liquid staking tokens (stETH, rETH)
• **MEV market:** up to 90% of Ethereum blocks pass through MEV-Boost - the block-building market generates hundreds of millions of dollars annually for validators and builders

The Merge: Seven Years in the Making

The transition to proof-of-stake was part of Ethereum's roadmap from the very beginning in 2015. The difficulty bomb - an exponential difficulty increase mechanism - was built into the protocol as a "deadline" forcing the transition. But reality proved more complicated: the difficulty bomb was delayed **6 times**. In 2016, a research team led by **Vlad Zamfir** developed Casper - a PoS protocol for Ethereum. Work on sharding proceeded in parallel. By 2019 it became clear that combining The Merge and sharding was too risky - they decided to separate them. **Danny Ryan** led the Beacon Chain specification, which launched on December 1, 2020 with a genesis event involving 21,063 validators. **Tim Beiko** coordinated the transition through a series of testnets - Kiln, Ropsten, Sepolia, Goerli - each going through The Merge before mainnet. On September 15, 2022 at 06:42:42 UTC, the PoW chain reached the Terminal Total Difficulty, and block #15,537,394 was proposed by a PoS validator. Vitalik tweeted: "And we finalized!" At the Ethereum Foundation in Berlin, developers greeted the moment with applause - after seven years of work.

The Merge is the largest live blockchain migration in history. It proved that the fundamental consensus mechanism of a live network with millions of users and hundreds of billions of dollars can be changed without disrupting a single application.

Предварительные знания

• Proof of Stake
• EVM: Ethereum Virtual Machine

Beacon Chain: The Coordination Layer

On December 1, 2020, the **Beacon Chain** launched - a separate blockchain running in parallel with Ethereum's main PoW chain. Its job was to coordinate validators in the upcoming proof-of-stake system. For nearly two years the Beacon Chain ran "idle": validators staked ETH and earned rewards, but **not a single user transaction** passed through it.

Why launch a system two years before using it? It was a **live stress test**. Billions of dollars were at stake - you cannot switch the consensus of a $200B+ network without real-world testing. By the time of The Merge, the Beacon Chain had more than **400,000 validators** with a combined stake exceeding 13 million ETH (~$20B).

The Beacon Chain introduced a new time structure. Instead of the arbitrary PoW block time (average ~13 sec with large variance), it introduced **fixed 12-second slots**. 32 slots form an **epoch** (~6.4 minutes). Each slot is assigned one proposer (who proposes the block) and a committee of attesters (who vote for the block).

**Finalization in 2 epochs.** A block is considered final after ~12.8 minutes (2 epochs). Once finalized it cannot be reverted without destroying 1/3 of all stake. This is a much stronger guarantee than "6 confirmations" in Bitcoin (~60 minutes), where a theoretical reorg is always possible with enough hash rate.

**The square root in the formula** is a key design element. The more validators, the **lower** each one's reward, but the **total** issuance grows as a square root. This creates balance: enough incentive for security without runaway inflation.

Execution Layer: Where Transactions Live

After The Merge, the old PoW Ethereum chain did not disappear - it became the **Execution Layer** (EL). It is the same EVM, the same state, the same smart contracts. For dApps and users **nothing changed**: JSON-RPC API, addresses, balances, contracts - everything stayed in place. The only thing that disappeared was mining.

The Merge happened on **September 15, 2022** at 06:42:42 UTC. The switchover point was not a block number but the **Terminal Total Difficulty (TTD) = 58,750,000,000,000,000,000,000**. When the cumulative difficulty of the PoW chain reached this value, the last PoW block was mined, and the next block was proposed by a PoS validator. The network continued without a single second of downtime.

Architecturally, the Execution Layer handles everything related to **computation and state**: executing transactions in the EVM, updating the state trie, forming the execution payload (list of transactions + state root). EL clients are **Geth**, **Nethermind**, **Besu**, **Erigon**, **Reth**.

**Engine API** is the internal bridge between the consensus and execution clients. The Consensus Layer tells the execution layer: "here are the parameters for the next block - assemble transactions and execute them". The Engine API uses JSON-RPC but is **not accessible externally** - it is a private channel between two processes on the same node.

Consensus Layer: Clients and Diversity

The **Consensus Layer** (CL) is the "brain" of Ethereum after The Merge. It decides which block is canonical (fork choice rule - LMD-GHOST + Casper FFG), coordinates validators, manages staking and finalization. The CL operates through the **Beacon API** - a new REST API standardized for all clients.

Every full Ethereum node after The Merge consists of **two clients**: a consensus client + an execution client. This is not a coincidence - it is a deliberate architectural choice. Two independent codebases, two processes, two sets of developers. If a critical bug is found in one client - the other keeps working.

**Client diversity is a matter of network survival.** In 2022, Prysm controlled **~66% of the Beacon Chain** - above the 2/3 threshold needed for finalization. If Prysm had a consensus-breaking bug: 1. 66%+ of validators would create an **invalid chain** 2. That chain would be **finalized** (66% > 2/3 threshold) 3. Other clients would reject it as invalid 4. **Two finalized chains** would exist - a catastrophe requiring manual intervention 5. Validators on Prysm would lose stake through **slashing** for finalizing invalid blocks The community is actively working to reduce any single client's dominance below 33%.

In addition to proposing and attesting blocks, the CL manages **sync committees** - groups of 512 validators that sign every block for ~27 hours. Sync committees allow light clients (mobile wallets, browsers) to verify blocks without downloading the entire chain - checking 512 signatures is sufficient.

MEV-Boost: Separating Proposer and Builder

After The Merge, validators became block proposers. They gained power: **choosing** which transactions to include in a block and in what **order**. This created a market for **MEV** (Maximal Extractable Value) - value that can be extracted by manipulating transaction ordering. Arbitrage, liquidations, sandwich attacks - all of this is MEV.

The problem: if every validator optimizes MEV themselves, it leads to **centralization**. Large operators with sophisticated algorithms earn more → attract more stake → get more slots. Small solo validators lose out.

The solution: **Proposer-Builder Separation (PBS)**. The idea is to separate roles. A **builder** is a specialized agent that assembles the most profitable block from mempool transactions. A **proposer** is a validator that selects the best block from those offered by builders. MEV-Boost from Flashbots implements PBS as an external sidecar process, until PBS is enshrined in the protocol.

At its peak, **~90% of all Ethereum blocks** were built via MEV-Boost. This means the vast majority of validators prefer to receive a guaranteed builder bid rather than assemble a block themselves. For a solo validator with 32 ETH, MEV-Boost is the only way to compete with large operators.

**Censorship problem: OFAC compliance.** In August 2022, OFAC (US Treasury) added Tornado Cash to the sanctions list. Some relays (primarily Flashbots) began **filtering transactions** linked to Tornado Cash. In November 2022, up to 78% of blocks excluded these transactions. This called into question **censorship resistance** - a fundamental property of Ethereum. If a relay can censor transactions, and 90% of blocks go through relays, then the network is de facto censorable. Community response: Flashbots open-sourced the relay code, **non-censoring relays** appeared, and Tornado Cash transactions were included in blocks from the remaining ~22% of "compliant" validators. Today the **inclusion delay** for sanctioned transactions is a few blocks, not a total block.

**Enshrined PBS (ePBS)** - the plan to bake PBS directly into the Ethereum protocol, removing the dependency on external relays. This will eliminate trust in relays and make the block auction part of consensus. While ePBS is in development, MEV-Boost remains the de-facto standard.

Key Ideas

• **Beacon Chain** launched December 1, 2020 and ran in parallel with PoW for nearly two years as a live stress test. Slots (12 sec), epochs (32 slots), RANDAO for role assignment, finalization in 2 epochs (~12.8 min)
• **Execution Layer** - the former PoW chain, which after The Merge preserved the EVM, state, contracts, and JSON-RPC API. The transition was fully transparent to dApps. The Engine API connects the EL to the Consensus Layer
• **Consensus Layer** manages fork choice (LMD-GHOST + Casper FFG), validators, and finalization. Client diversity is critical: if one client exceeds 66%, a bug in it can finalize an invalid chain
• **MEV-Boost** implements Proposer-Builder Separation: builders assemble blocks, a relay validates them, the proposer picks the best bid. Up to 90% of blocks are built via MEV-Boost, but censorship concerns arise (OFAC/Tornado Cash)
• The Merge only changed the consensus - the "plane engine in mid-flight" from the lesson opening. Lower fees and scalability are tasks for separate upgrades (The Surge), not The Merge

Related Topics

The Merge connects consensus, cryptography, and Ethereum's economics into a unified architecture:

• Proof of Stake — The Merge activated PoS as the primary consensus mechanism - the Beacon Chain is an implementation of the principles covered in the PoS lesson
• Gasper: LMD-GHOST + Casper FFG — The Consensus Layer uses Gasper for fork choice and finalization - Ethereum's specific PoS protocol
• BLS Signatures — Attestations, sync committees, and aggregated signatures in the Beacon Chain use BLS - the cryptographic primitive enabling efficiency with 900K+ validators
• Ethereum Roadmap — The Merge is the first stage. Next: The Surge (sharding), The Scourge (MEV), The Verge (Verkle trees), The Purge (state expiry), The Splurge

Вопросы для размышления

• Ethereum chose a two-year parallel Beacon Chain launch before The Merge. What alternative migration strategies could have been used, and why did the team reject faster options?
• MEV-Boost centralized block building: a few builders and relays control most blocks. Does Enshrined PBS solve this problem, or does it simply move centralization to another level?
• If Prysm had contained a consensus-breaking bug during its 66% dominance, what would the real network recovery scenario have looked like? How much ETH would have been lost to slashing?

Связанные уроки

• dist-10-byzantine