TrueStake
THE MERGE · 15 SEP 2022

Proof of Stake

A way to secure a network with capital, not electricity — and it created a new economy.

Proof of Stake went live on Ethereum on 15 Sep 2022 — an event called the Merge.

~99.99%less energy
vs. Ethereum’s prior Proof-of-Work era (CCRI)
01 / PROOF OF STAKE

The big idea in 90 seconds

Two ways to secure a blockchain — burn electricity, or stake capital.

Every blockchain needs a way to decide who adds the next block, and to make it expensive to cheat. There are two ways to do that: burn electricity — Proof of Work, the way Bitcoin still secures itself — or stake capital — Proof of Stake. Ethereum launched on Proof of Work, then switched to Proof of Stake.

WHAT CHANGED
  • Mining was replaced by validator-based block production.
  • Energy use fell by ~99.99%.
  • New ETH issuance now flows entirely through staking yield — never mining.
WHAT DIDN’T CHANGE
  • Gas fees did not drop.
  • No transaction history was wiped — the full chain history was preserved.
  • ETH is still just ETH — no new token, no swap.
  • There is no "ETH2" — that name was retired, not a separate chain or asset.

Source: ethereum.org, “The Merge.”

02 / PROOF OF STAKE

How Proof of Stake launched

Proof of Stake didn't arrive in a single moment — it emerged over nearly two years, then went live in one named event.

The Beacon Chain launched on 1 Dec 2020 and ran Proof of Stake in parallel with Ethereum’s still-live Proof-of-Work mainnet for nearly 21 months, accumulating real validator history before the two chains switched over. Three public testnets rehearsed the exact transition mechanics that summer. Then, on 15 Sep 2022, Proof of Stake went live — an event called the Merge. Staking kept evolving afterward: withdrawals, compounding validators, and scaling upgrades all shipped in the years since.

1 Dec 2020
Beacon Chain genesis
PoS runs in parallel · 21,063 validators
Jun–Aug 2022
Testnet merges
Ropsten · Sepolia · Goerli — dress rehearsals
6 Sep 2022
Bellatrix
consensus-layer prep
THE MERGE
15 Sep 2022
Proof of Stake goes live
Paris upgrade · Terminal Total Difficulty · block 15,537,394
12 Apr 2023
Shapella
staking withdrawals enabled for the first time
7 May 2025
Pectra
compounding validators · EL-triggered exits
3 Dec 2025
Fusaka
scaling · PeerDAS

The Merge itself was two forks, ~9 days apart: Bellatrix (6 Sep 2022) prepared the consensus layer, then the Merge (Paris, 15 Sep 2022) flipped the switch at Terminal Total Difficulty — execution-layer block 15,537,394. Withdrawals didn’t exist yet; Shapella (12 Apr 2023) made staking spendable for the first time. Pectra (7 May 2025) and Fusaka (3 Dec 2025) kept the network improving.

Go deeper: why upgrade names come in pairs

Ethereum upgrades a consensus-layer client and an execution-layer client separately, so a single upgrade often carries two names, one per layer: the Merge paired Bellatrix (consensus) with Paris (execution). The same pattern produced Shapella — a portmanteau of Shanghai (execution) and Capella (consensus).

For every fork in order — including the ones before and after this launch story — see the Ethereum upgrade timeline.

03 / PROOF OF STAKE

Proof of Work vs Proof of Stake

Two different answers to the same question: who adds the next block, and what stops them from cheating?

Under Proof of Work, miners spend electricity racing to solve a computational puzzle — hashpower burned in trial and error, with a difficulty target that adjusts to keep the race taking roughly the same amount of time. Whoever finds a valid hash first proposes the next block. Security comes from a sunk cost: the hardware and electricity are already spent, reusable for the next race regardless of how the miner behaves.

Under Proof of Stake, validators bond 32 ETH of capital to participate. Each 12-second slot, one validator is randomly selected to propose a block, and a randomly-assigned committee casts attestations — votes on whether that block is valid. Thirty-two of those slots make an epoch, roughly 6.4 minutes. Security comes from capital at risk: misbehave — propose two conflicting blocks, or submit contradictory attestations — and that stake is slashed, destroyed outright, not merely idled.

MINERS
Hashpower
spend electricity racing to solve a puzzle
Difficulty
target adjusts to hold the race's pace
WINNER PROPOSES A BLOCK
First valid hash wins
broadcasts the new block
Block added (PoW)
finalityprobabilistic
conventionmore confirmations = more confidence
assurancedeeper burial = safer
no fixed cutoff — never absolute
VALIDATORS
32 ETH stake
bonded capital, put at risk to participate
Committee
random each slot (12s) · epoch = 32 slots ≈ 6.4 min
COMMITTEE ATTESTS
Attestations
votes on the proposed block's validity
Misbehave → slashed
equivocation or contradictory votes destroy stake
Block added (PoS)
finalityCasper FFG checkpoint
threshold≥⅔ of ALL staked ETH
time~2 epochs (~12.8 min)
finalized = permanently irreversible

Both mechanisms end the same way — a block gets added — but how sure you can be it stays added is where they diverge. Proof of Work's finality is probabilistic: each additional block mined on top raises confidence, but no fixed number ever makes it absolute. Proof of Stake's finality is economic and checkpointed — Ethereum’s Casper FFG requires votes representing at least two-thirds of all staked ETH, twice in a row, to make a block finalized: permanently irreversible, in roughly two epochs — about 12.8 minutes.

Go deeper: what actually gets you slashed

Three behaviors are slashable, all forms of equivocation: double-vote (two attestations for the same target epoch but different chain histories), surround-vote (an attestation that contradicts an earlier one by “surrounding” its span), and double-proposal (proposing two different blocks for the same slot).

The penalty is an immediate minimum deduction plus a correlated-slashing component that scales with how many other validators are slashed around the same time — up to the validator’s entire stake if many are slashed together — plus a forced exit from the validator set.

Go deeper: why finality is economic, not just probabilistic

Rewriting a block that’s already finalized isn’t a matter of out-computing the network, the way reorganizing a Proof-of-Work chain is. It requires validators controlling roughly a third or more of all staked ETH to attest to two conflicting histories at once — and doing that is a double-vote, one of the exact behaviors slashing exists to punish.

That means an attacker attempting to reverse a finalized block doesn’t just need a lot of capital — they destroy a third or more of all staked ETH in the process of trying. Finality isn’t a probability that fades toward certainty; it’s a threshold backed by capital that gets burned the moment it’s crossed dishonestly.

Go deeper: the attack-threshold ladder
  • ~33% of stake, offline or attesting maliciously — a two-thirds supermajority can no longer form, so the chain stops finalizing new blocks (a liveness failure, not a takeover). The inactivity leak gradually reduces non-participating validators’ stake until they fall back under the threshold.
  • 34% of stake — enough to force two conflicting checkpoints to both finalize at once, but only by double-voting — a slashable offense that destroys the attacker’s entire equivocating stake in the act.
  • 51% (a simple majority) of stake — enough to dominate the fork-choice rule: short reorgs, censoring specific transactions, and extracting maximal value by reordering blocks. The Ethereum community can still route around this by continuing to build on the honest chain.
  • 66% (two-thirds) of stake — full control over finality itself: an attacker at this threshold can finalize whatever chain it prefers outright — spend, rewind, spend again, censor transactions, and reorg at will.
04 / PROOF OF STAKE

Why Proof of Stake?

The honest case for staking capital instead of burning electricity — six benefits, and what each one costs.

ENERGY
~99.99% less energy

Proof of Stake secures Ethereum with staked capital instead of burned electricity, cutting the network's energy draw by roughly 99.99% overnight.

Go deeper: how confident is that number?

The reduction itself isn’t in question — multiple independent methodologies land on a cut of this order of magnitude. What’s contested is the precise pre-Merge baseline, since the two leading estimates measure it differently. The full methodology note (CCRI vs. Digiconomist) is below, in the energy story.

DECENTRALIZATION
No mining rig required

Running a validator takes modest, consumer-grade hardware — no ASIC farm or GPU rig racing for hashpower.

Go deeper: the 32 ETH bar is real, and it's a different kind of barrier

Hardware isn’t the barrier anymore, but capital is: bonding 32 ETH to run a solo validator is a genuine, purely financial barrier that has nothing to do with computing power. Smaller stakers reach the network through Hosted staking, pools, or Liquid Staking instead.

And the tools that make participation easier have themselves grown concentrated: a single Liquid Staking provider has at times held a large enough share of all staked ETH to draw real centralization concern — a pressure the hardware story alone doesn’t capture.

ECONOMIC SECURITY
Attacking it destroys your own capital

The cost of attacking the network scales with ETH's price and the stake required, and misbehaving validators get their stake slashed — destroyed, not just idled like unused mining hardware.

Go deeper: security priced in a currency that moves

Because attack cost is denominated in ETH’s own market price, the network’s security budget rises and falls with that price — a sharp price drop cheapens an attack in dollar terms, even though the mechanism (slashing capital) is unchanged. This is a description of a mechanism, not a fixed dollar figure for “how much an attack costs today.”

RELIABILITY
A clean track record since the Merge

Proof of Stake has run with long slashing-free stretches since going live in September 2022 — a real-world reliability record, not just a design promise.

Go deeper: it's still a short history

That track record spans a few years, not decades — a far shorter sample than Bitcoin’s Proof-of-Work run. “Reliable so far” is an honest claim; it isn’t a guarantee that no future stress test will find a gap.

CREDIBLE NEUTRALITY
Anyone can participate — the network doesn't pick winners

Staking is permissionless: any validator that bonds ETH and follows the rules can propose and attest to blocks. That's credible neutrality — no operator gets special treatment.

Go deeper: neutral at the protocol layer, less so at the block-building layer

Validator selection itself is protocol-neutral, but a separate market — the builders and relays that assemble many of today’s blocks — has at times excluded specific transactions (for example, OFAC-compliant relays), a centralization pressure the roadmap’s enshrined-PBS work aims to close.

LOWER ISSUANCE
New ETH issuance fell roughly 88%

Proof of Stake issues roughly ~0.5% of the network's ETH supply per year to validators — versus the ~4% that went to miners under Proof of Work.

Go deeper: a network average, not a promise

That issuance figure is a network-wide average, and per-validator yield trends down as more ETH stakes — it isn’t a fixed rate any one staker is guaranteed, and it says nothing about execution-layer tips or MEV, covered later on this page.

05 / PROOF OF STAKE

The energy story

A ~99.99% cut in the electricity it takes to secure Ethereum — verified from actual measurements, not just a design promise.

Before the Merge, securing Ethereum by mining took real, measurable electricity — roughly 21–23 TWh every year, on the same order of magnitude as Bitcoin, global gold mining, or the world’s data centers. After the Merge, that draw collapsed to roughly 0.0026 TWh/yr — smaller than the electricity a single major streaming service, like Netflix, uses to run its own operations. The figure ethereum.org treats as primary comes from CCRI (Crypto Carbon Ratings Institute), a bottom-up study built from actual hardware measurements, not an inference from prices: a cut of about 99.99% — more precisely, over 99.988% of electricity and roughly 99.992% of the network’s carbon footprint.

PROOF OF WORK
~21–23 TWh/yr
≈ same order as Bitcoin · gold mining · global data centers
PROOF OF STAKE
~0.0026 TWh/yr
smaller than a single streaming service's own footprint
Go deeper: the methodology gap (CCRI vs. Digiconomist)

CCRI’s bottom-up, hardware-measurement study is the estimate ethereum.org endorses as primary, putting Ethereum’s pre-Merge electricity draw at roughly 21–23 TWh/yr. Digiconomist’s competing top-down estimate — inferred from miner-revenue economics rather than measured hardware — puts pre-Merge Ethereum meaningfully higher; its published series has ranged roughly 28–112 TWh/yr over time.

That gap is a genuine methodology difference — bottom-up hardware measurement vs. top-down economic inference — not a disagreement that the Merge happened or roughly how large the cut was. Both approaches land on a reduction of about the same order of magnitude, ~99.99%. This page cites CCRI as primary and states the gap plainly rather than resolving it as settled.

Source: CCRI, “The Merge — Implications on the Electricity Consumption and Carbon Footprint of the Ethereum Network” (Sept 2022); ethereum.org, “Ethereum energy consumption.”

06 / PROOF OF STAKE

A new economy is born

Proof of Stake pays a yield for securing the network — and that single fact spawned an entire industry.

That single mechanical fact — a validator earns a yield for helping secure the network — is what spawned every identity you see below. Some stakers run their own validator infrastructure directly: a Solo staker runs their own hardware, a Hosted staker owns the validator but delegates node operation, and a Pool operator runs a node inside a permissionless pool. That's Validator Staking. Others hold a token that represents staked ETH instead: a Liquid staker holds a Liquid Staking Token, and a Liquid restaker takes that a step further, reusing the same staked capital to also help secure other services in exchange for extra yield. That's Liquid Staking.

A whole economy of participants, all incentivized to keep the network honest. That isn’t a side effect of Proof of Stake — it’s the security model itself, expressed as economics.

SOLO STAKER
runs their own validator hardware
HOSTED STAKER
delegates node operation, keeps the keys
POOL OPERATOR
runs a node inside a permissionless pool
LIQUID STAKER
holds a Liquid Staking Token
LIQUID RESTAKER
also secures other services
YIELD
one number
not one clean source
assembled from many sources — unpacked next

This industry-wide phenomenon — network yield, how much ETH is staked, issuance — is what this page calls Staking Economics: the field a reader is entering just by learning this page. It is not what TrueStake sells. The per-customer record TrueStake reconciles, to the wei, is a separate thing entirely — Staking Financials, covered later on this page. The contrast is clean: the field you’re entering, versus your own books.

And here’s the hook: that yield is not one clean number. It’s assembled from several different sources at once — the next section unpacks where they come from — and accounting for it precisely, down to the wei, is genuinely hard. That difficulty is the outcome Proof of Stake created.

Go deeper: why two terms, not one?

Staking Economics is analytical and systemic — it describes the forces, incentives, and returns that govern ETH staking as an industry. It answers “is staking worthwhile, and how does the field perform?” This page educates about that field and benchmarks against it, but doesn’t sell it.

Staking Financials is concrete and backward-looking — it’s an actual record of what happened: what a specific staker’s validators actually earned, provably, reconciled to the wei against on-chain receipts. That’s what TrueStake builds. One word describes a field; the other describes a set of books. Using one term for both was the confusion this page is built to avoid.

Source: ethereum.org, “Proof of Stake” and “Rewards and penalties”; CONTEXT.md staking-tier and identity vocabulary; ADR-0151.

07 / PROOF OF STAKE

The economics up close

Issuance, the burn, and where a validator's yield actually comes from — the substance behind the thesis, at whatever depth you want it.

The Merge didn’t just change how blocks get made — it collapsed how much new ETH the protocol pays out. Before the Merge, miners earned roughly 13,000 ETH a day in block issuance; the day Proof of Stake went live, that fell to about 1,600 ETH a day paid to validators instead — an overnight drop of roughly 88%, from an annual issuance rate of about 4% down to about 0.5%. It’s sometimes called the “triple halving” — the cut is roughly three Bitcoin halvings’ worth of supply reduction, landing in a single day.

Separately — and this is worth stating precisely, since the two get conflated — Ethereum has burned the base fee of every transaction since EIP-1559 (the London upgrade) went live in August 2021, more than a year before the Merge. Net ETH supply change is just issuance minus that burn: when the burn outpaces issuance, supply shrinks — which is where the phrase “ultrasound money” comes from. Ethereum genuinely was net-deflationary for about 18 months after the Merge. But that’s not the whole story: it is not cumulatively deflationary since the Merge — total supply is higher today than it was on 15 Sep 2022, because the Dencun upgrade (Mar 2024) made posting Layer-2 data to Ethereum dramatically cheaper, pulling fee-paying activity off mainnet and collapsing the burn. Ultrasound money is a demand-dependent thesis, not a guaranteed protocol property — check the live number at ultrasound.money rather than trusting a figure frozen on this page.

ISSUANCE
~1,600 ETH/day
~0.5%/yr — down ~88% from ~13,000 ETH/day pre-Merge (~4%/yr)
sometimes called the "triple halving"
BURN (EIP-1559)
base fee, every transaction
London, Aug 2021
predates the Merge by more than a year — demand-dependent, not fixed
NET SUPPLY CHANGE
issuance − burn
"ultrasound money": demand-dependent thesis, not a guaranteed property
not cumulatively deflationary since the Merge — see live
CONSENSUS-LAYER
attestations · sync committee · proposals
steady — earned every epoch by every active validator
EXECUTION-LAYER + MEV
priority tips + MEV
spiky — earned only when a validator proposes a block
YIELD
one yield, many sources
consensus = the large majority, network-wide — a cited average, never "your yield"
Solo: full yield · Liquid Staking: ~10% protocol fee (e.g. Lido)

That net-issuance number is the supply-wide picture. Zoom into a single validator, and its yield is assembled from two different layers. A steady consensus-layer stream — attestations, sync-committee duty, block proposals — pays every active validator, every epoch, just for doing its job correctly. A spikier execution-layer stream — priority tips plus MEV — only pays when a validator is actually selected to propose a block, which is rare for any one validator. Consensus is the large majority of network-wide yield most of the time; where a source cites a split at all, it’s a network-average Ethereum fact — never a claim about any specific staker’s own split. Blended yield network-wide has held in a low single-digit range since the Merge, and it trends down as more ETH stakes (more on why, below). A Solo staker captures that full yield directly; Liquid Staking (Lido, the largest provider) takes a protocol fee of about 10% off the top.

Go deeper: why does yield fall as more ETH stakes?

Each validator’s base_reward is proportional to its effective balance and inversely proportional to the square root of total active stake network-wide. Because total issuance grows only as fast as √(total staked) while each validator’s own share of it shrinks as 1/√(total staked), blended network yield falls smoothly every time more ETH joins the validator set — and rises if ETH exits.

That’s a deliberate design choice, not an accident: Ethereum wants the incentive to self-balance. A validator set that’s already large has less reason to keep growing further; a validator set that shrinks has more reason to (re)join. It’s the same mechanism that took network yield from over 15% at Beacon Chain genesis (Dec 2020, a tiny validator set) down to the low single digits today, as the validator set crossed a million.

The full breakdown of what a validator earns each epoch — the nine separate consensus-layer components (attestation source, target, and head; finality-delay penalty; sync-committee duty; proposer-inclusion bonuses; and more) plus how execution-layer MEV gets attributed on-chain — lives on the Ethereum consensus standard page. This page deliberately doesn’t reproduce that detail.

Source: ethereum.org, “How the Merge Impacted ETH Issuance” and “Proof-of-stake rewards and penalties”; EIP-1559; ultrasound.money (live); Lido, “Protocol Fee”; ADR-0110.

08 / PROOF OF STAKE

Staking today & tomorrow

The system didn't stop at the Merge. It keeps shipping — and it's spawned an entire second layer of yield-stacking, extra-risk economics on top.

Proof of Stake has kept improving every year since launch, and each upgrade is shipped, not a promise. Shapella (Apr 2023) turned on withdrawals for the first time — staked ETH could finally leave the protocol. Pectra (May 2025) shipped MaxEB: compounding validators, whose effective-balance cap jumped from 32 ETH to 2,048 ETH (EIP-7251), plus execution-layer-triggered exits (EIP-7002) — an operator can now initiate an exit from their own wallet instead of only from the validator client. Fusaka (Dec 2025) shipped PeerDAS (EIP-7594), scaling how much data the network can verify per slot.

Underneath that shipped history sits a real tension. Some stakers run Validator Staking directly (Solo · Hosted · Pool); many more hold a token instead — Liquid Staking. Liquid Staking is enormously convenient, but it concentrates: one provider has historically represented roughly a quarter to a third of all staked ETH — check the current split at DefiLlama rather than trusting a number frozen on this page. That's the honest downside of convenience: a large share of the validator set following one operator's decisions. Distributed Validator Technology (DVT) — splitting a single validator's signing key across multiple independent node operators — is the industry's answer: it lets a pool decentralize who actually runs the hardware behind each validator, without asking any staker to give up Liquid Staking's convenience.

Restaking pushes that same idea further. EigenLayer let a Liquid restaker reuse already-staked ETH — or the Liquid Staking Token wrapping it — to also help secure additional external services, earning extra yield for taking on extra slashing conditions on top of ordinary consensus slashing. ether.fi’s eETH is the largest Liquid Restaking Token (LRT) built on this pattern. This is a genuine appeal — more yield from the same capital — but the risk is equally real and no longer theoretical: EigenLayer’s slashing mechanism went live on mainnet in April 2025, and restaking capital proved anything but sticky once real downside risk activated. Opting into multiple services stacks compounded, correlated slashing exposure — more yield sources, more independent ways to lose principal, not a free multiplier. The underlying capital securing more than one system at once is a genuine new form of rehypothecation-style leverage. A small number of large operators concentrate a large share of restaked capital — the same concentration concern as Liquid Staking, one layer deeper — and every additional contract in the stack (the restaking protocol, the LRT, each secured service) is separately exploitable smart-contract risk. Restaking TVL moves fast; check it live at DefiLlama rather than trusting a figure frozen here.

SHAPELLA
12 Apr 2023
Withdrawals go live
first-ever exit for staked ETH
SHIPPED
PECTRA
7 May 2025
MaxEB compounding
32 → 2,048 ETH · EIP-7251/7002
SHIPPED
FUSAKA
3 Dec 2025
PeerDAS scaling
EIP-7594
SHIPPED
GLAMSTERDAM
H2 2026 (est.)
Enshrined PBS
EIP-7732 · builder on-protocol
SCHEDULED
SINGLE-SLOT FINALITY
Not scheduled
Years out
no upgrade assigned yet
RESEARCH-STAGE
VERKLE TREES
Not scheduled
Deprioritized
uncertain beyond Glamsterdam
RESEARCH-STAGE
LOWER THE MINIMUM
Not scheduled
Discussed only
not an assigned upgrade
RESEARCH-STAGE

What comes next is real, but it isn’t all equally close. Enshrined PBS (ePBS, EIP-7732) is scheduled for Glamsterdam, targeting the second half of 2026 — no mainnet date is set yet, so treat that as an estimate, not a promise. It would move today’s off-protocol MEV-Boost relay market on-protocol, separating who builds a block from who proposes it as a native consensus rule (more on why that matters in the next section). Beyond that, three ideas are genuinely research-stage — years out, not scheduled: single-slot finality, lowering the 32 ETH minimum needed to run a validator, and Verkle trees / statelessness (which has been deprioritized, and is uncertain beyond Glamsterdam). They’re real research directions, not vaporware — but naming them as research, not roadmap, is exactly the honesty this page is built around.

Go deeper: restaking's risk math, and the Surge/Scourge framing

Why restaking risk compounds, not just adds. A Liquid restaker who opts a single position into three different services doesn’t take on “three times one risk” — they take on three independent slashing conditions, any one of which can fire on its own. That’s structurally similar to how rehypothecated collateral works in traditional finance (the same underlying asset backing more than one obligation at once), without claiming the risk profiles are identical. EigenLayer’s own TVL fell from a peak above $15B to roughly $7B by late 2025 following its April 2025 slashing launch — a dated, real data point that restaking capital moves fast once downside risk is no longer theoretical.

Where TrueStake stands today. TrueStake already covers the Liquid Staking side of a Liquid restaker’s position — the underlying consensus yield the LST layer earns. The additional AVS-specific yield restaking itself creates is explicitly out of today’s scope.

The “Surge / Scourge / Verge / Purge / Splurge” framing. You may see Ethereum’s roadmap described through this six-phase mental model — Vitalik Buterin’s own name for the categories of work still ahead (scaling, MEV/censorship-resistance, statelessness, history expiry, and everything else). It’s a useful narrative lens and still widely used as an umbrella for related upgrades, but ethereum.org’s own roadmap page is organized around the chronological fork list above, not these six names — so this page leads with fork names and treats the six-phase framing as a secondary, narrative-only label.

Source: ethereum.org roadmap; EIP-7251, EIP-7002, EIP-7594, EIP-7732; EigenLayer documentation; ether.fi; DefiLlama (live); the eth-consensus-data skill.

09 / PROOF OF STAKE

The MEV challenge

The honest messy part of the story — where extra value comes from at the execution layer, and why it's a centralization and censorship concern, not just an extraction one.

Not everything a validator earns comes from simply doing its job correctly. Maximal Extractable Value (MEV) is value that comes from reordering, including, or excluding transactions within a block — from choosing whose transaction goes where, not from validating anything. It shows up in a few recognizable forms: DEX arbitrage (correcting price differences across exchanges) and liquidations (enforcing a lending protocol’s own rules) are largely benign side effects of an efficient market. Sandwich attacks are not — they front-run and back-run an ordinary trader’s own order to profit directly at that trader’s expense. MEV isn’t a bug; it’s a structural consequence of whoever controls block construction having discretion over transaction order.

Here’s how it actually flows today. Searchers — permissionless bots, anyone can run one — scan the mempool for MEV opportunities and bid for them. Builders compete to assemble the single most valuable block they can from those bids plus ordinary transactions, then bid to pay a validator for the right to have that block proposed. Relays sit in between as trusted intermediaries: they check a builder’s block is valid and pays what it claims, then hand the proposer only a header — not the block’s actual contents — so the validator can sign it blind, without either side having to trust the other with an unpaid, unverified block. This is MEV-Boost, built on Proposer-Builder Separation (PBS): separating who builds a block from who proposes it. It’s the same spiky, proposal-only execution-layer stream the previous section’s yield composition described — most of a validator’s yield is the steady consensus-layer stream; MEV is additive on top of it, and only on the rare slot a validator actually proposes.

SEARCHERS
arbitrage · sandwich attacks · liquidations
sandwich attacks harm ordinary users
many, permissionless — anyone can run one
BUILDERS
assembles the most valuable block it can
competes on payment to the proposer
a small number of builders build most blocks
RELAYS
passes the winning block to the proposer
sight-unseen — a trusted middleman
some relays filter sanctioned transactions — others don't
PROPOSER
signs the header it never inspected
one proposer, per slot
today's safeguard: MEV-Boost / PBS, below

The real concern isn’t extraction itself — it’s what centralizes around it. A small number of builders build most blocks, concentrating real discretion over what does and doesn’t make it into the chain into a few hands. And this is also exactly where censorship happens in practice: some relays filter out OFAC-sanctioned transactions before relaying a block; others don’t. Which relays dominate volume — and so what share of Ethereum blocks pass through a censoring relay — has swung widely over time; a number printed on this page today would be stale within months. Check the current, live share at mevwatch.info rather than trusting a figure frozen here.

The roadmap answer is to move that separation on-protocol: enshrined PBS (ePBS, EIP-7732), scheduled for the Glamsterdam upgrade — but treat that as a direction, not a settled fact. ePBS is a confirmed fork headliner, yet as of this writing it (and a related upgrade, EIP-7805) remain formally in Draft status, and its design has kept moving even recently — the debate is real, not unanimous. ePBS also isn’t the whole censorship-resistance story: a separate mechanism, FOCIL (Fork-Choice-enforced Inclusion Lists), targets censorship resistance specifically rather than the builder/proposer trust problem ePBS solves — and it’s deferred to Hegotá, the fork after next.

Go deeper: the relay/builder market and censorship history

The relay market is a mix, not a monolith. At any given time, some active relays filter OFAC-sanctioned transactions before relaying a block and some don’t — and which relays carry the most volume changes as operators enter, exit, and adjust policy. Treat any specific relay name or market-share figure as a snapshot, not a durable fact; relayscan.io and mevwatch.info track this live.

Censorship share has swung widely, not settled into a steady state. The compliant (censoring) share of relayed blocks has moved substantially since OFAC first sanctioned Tornado Cash addresses in 2022 — rising, then falling back — driven by proposer preference, new non-filtering relays entering the market, and public pressure on operators. That volatility is itself the reason this page links live rather than asserting a number.

ePBS doesn’t directly fix censorship — FOCIL does. Enshrined PBS moves today’s off-protocol MEV-Boost relay market on-protocol, but its job is trust between builder and proposer, not guaranteeing inclusion. The mechanism actually aimed at censorship resistance — Fork-Choice-enforced Inclusion Lists (FOCIL, EIP-7805) — is a separate, later piece of the roadmap: it’s Hegotá’s confirmed headliner, one fork after Glamsterdam, not bundled with ePBS.

Source: ethereum.org, “Maximal extractable value (MEV)”; Flashbots documentation; the eth-consensus-data skill (MEV-Boost flow, relay landscape); mevwatch.info (live); EIP-7732; EIP-7805; ADR-0110.

10 / ACCOUNTING

Proof of Stake made this Income genuinely hard to account for.

A staker’s real Income is assembled from consensus-layer yield, execution-layer tips, and MEV, minus penalties, across many validators and addresses — in gwei that must reconcile to on-chain receipts to the wei. That complexity is the outcome Proof of Stake created. Precisely accounting for it — across every validator type, every yield stream — is exactly TrueStake’s niche: ETH Staking Financials, the audit-defensible, per-customer record of what your staking actually earned, provably. That’s the difference this page has been building to: Staking Economics is the field you just learned about; Staking Financials is your own books.

get_started →

New to staking-income timing? Read when staking income is recognized.

Not tax advice · consult your tax advisor.
11 / PROOF OF STAKE

Keep learning

Where to go next.

SOURCES

Get the next piece

Staking tax · Policy · Yield decomposition. About one piece a month. No spam.

Proof of Stake — how Ethereum secures a network with capital, not electricity · TrueStake