Proof of stake is now how most blockchains reach consensus. But "validators stake ETH" doesn't explain what's actually happening. Why does locking up money make a transaction history trustworthy? The proof-of-work intuition—miners compete to solve puzzles—doesn't apply here. Staking isn't a competition. It's a mechanism where validators put capital at risk of permanent destruction in exchange for the right to participate in block production. Behave honestly, get paid. Cheat, lose your stake.
The security guarantee comes from economic skin in the game, not energy expenditure.
The Mechanism: From Deposit to Finalized Block
Here's how the process works on Ethereum, which runs the most-analyzed proof-of-stake implementation at scale.
Step 1: The validator deposit. A validator sends 32 ETH to a smart contract called the deposit contract. This locks the ETH as collateral. The validator can't spend it. They can't move it. Its only function is to back their behavior.
Step 2: Validator selection. The protocol selects a validator to propose the next block using a pseudo-random algorithm called RANDAO—a scheme where validators contribute entropy so no single party controls who gets selected next. Selection probability scales with stake size, though at 32 ETH per validator, most participants start at the minimum.
Step 3: Block proposal. The selected validator pulls pending transactions from the mempool, executes them through the Ethereum Virtual Machine, and broadcasts a proposed block to the network. This is where transactions actually get processed—the EVM runs each contract call, updates state, and calculates the new state root.
Step 4: Committee attestation. A randomly assembled committee of validators votes on whether to accept the proposed block. This process is called attestation. For the block to be included in the canonical chain, it needs a supermajority: at least two-thirds of the committee must attest in favor.
Step 5: Epoch processing. Every 32 blocks—roughly 6.4 minutes—is an epoch. At each epoch boundary, attestations are tallied, rewards are distributed, and penalties are applied. Validators who missed attestations or went offline receive small penalties. Validators who signed attestations get small rewards.
Step 6: Finality. The consensus layer runs two interacting mechanisms simultaneously. LMD-GHOST (Latest Message Driven Greedy Heaviest Observed SubTree) is the fork choice rule—it determines which chain the network treats as canonical moment to moment. On top of that runs Casper FFG, a finality gadget.
Casper FFG works in two-epoch cycles. After one epoch is "justified," and the next one is also justified, the first epoch becomes finalized. A finalized checkpoint is cryptographically irreversible: no honest validator can ever vote for a competing chain without facing slashing. At current Ethereum parameters, this means a transaction achieves finality in roughly 12-15 minutes after inclusion.
That's the full pipeline. Deposit → selection → propose → attest → epoch → finalize.
Where Constraints Live
The 32 ETH minimum is the most visible barrier. It's not arbitrary—it sets the cost basis for an attack. But it also prices out most retail participants from solo validation, which is why liquid staking protocols emerged. Lido, Rocket Pool, and others pool ETH from smaller holders, stake on their behalf, and return liquid tokens (stETH, rETH) representing the position.
This creates a structural tension. By late 2024, Lido controlled roughly 28% of all staked ETH. The community's red line is 33%, because a validator controlling one-third of the network can theoretically prevent finality—not by forging history, but by refusing to attest. This is a liveness failure, not a safety failure, but it's still a meaningful attack vector.
Slashing is the hard enforcement mechanism. Validators get slashed—a portion of their 32 ETH destroyed—for two specific behaviors: double voting (signing two different blocks at the same slot) and surround voting (signing an attestation that contradicts a prior one). These are detectable on-chain. The logic is automatic.
The security assumption is explicit: PoS requires at least two-thirds of validators to be honest. Break that assumption and the guarantee breaks.
What's Changing
The validator count on Ethereum expanded massively post-Merge—over 900,000 validators by late 2024. That's a lot of independent participants backing blocks, and it meaningfully improved decentralization at the validator level even as stake concentration through liquid staking remained a structural concern.
EigenLayer introduced restaking: validators can extend their staked ETH to simultaneously secure other protocols. The security of the original stake gets "rented" to AVSs (Actively Validated Services)—bridges, data availability layers, oracle networks. This expands the utility of staked ETH but adds complexity to the slashing surface. A validator participating in multiple AVSs faces slashing risk from multiple protocol layers, not just Ethereum consensus.
The broader shift is simpler: proof of stake is now the default. Ethereum's 2022 Merge was the decisive test. The chain transitioned from proof of work to proof of stake without downtime. Every major L1 that launched after 2022 defaulted to some variant of PoS.
Confirmation Signals
Watch these to assess whether the mechanism is operating as designed:
- Validator count and Nakamoto coefficient — more validators with lower concentration is healthier
- Attestation participation rate — above 95% indicates a well-functioning validator set
- Lido's share of staked ETH — any sustained move toward 33% warrants attention
- Finality time consistency — if epochs are regularly failing to finalize, something is wrong
- Slashing event frequency — rare is good; a spike could indicate a client bug
What Would Break This
The model fails under these conditions:
A cartel controls ≥⅓ of staked ETH and deliberately withholds attestations. This stops finality. Ethereum has a social layer response to this—inactivity leak, where offline validators lose stake until the honest majority regains supermajority—but it's a serious disruption.
A cartel controls ≥⅔ of staked ETH. Now they can finalize competing forks. This is the safety failure mode—the one the 32 ETH minimum is designed to make economically prohibitive.
MEV extraction becomes so profitable that reorging recent blocks is rational. This requires MEV opportunities large enough to exceed the slashing penalty plus expected future rewards from maintaining honest validator reputation. The numbers haven't supported this. None of these have materialized.
Timing Perspective
Now: Proof of stake is operational and stable. Ethereum has processed billions of transactions through this mechanism since the 2022 Merge without a finality failure.
Next (2025-2027): The active questions are validator economics under restaking complexity, the long-term equilibrium on liquid staking concentration, and whether EigenLayer's expanded slashing surface creates unexpected risks.
Later (2028+): Whether the stake concentration trajectory converges toward or diverges from the decentralization properties the mechanism was designed to produce. That answer isn't visible yet.
What This Doesn't Tell You
This explains how proof of stake validates transactions. It doesn't tell you whether proof of stake is more secure than proof of work—that's a different analysis. PoW converts energy into security; PoS converts capital into security. Both create economic deterrents against attacks through different mechanisms. Whether one design is "better" depends on which threat model you're optimizing against, and that's partly a values question.
It also doesn't say anything about whether ETH is a good investment. The mechanism working well and the asset appreciating are related but separate questions.
Epistemic status: Mechanism description based on documented Ethereum consensus layer specifications. Lido market share figures approximate as of 2024; check current dashboards for live data.
