What Is Blockchain Finality?

When a wallet says a transaction is “confirmed,” it is easy to assume the matter is closed. In many cases it effectively is, but confirmation and finality are not the same thing. Finality is the stronger property: the point at which a transaction cannot be undone. Different blockchains reach it in different ways and on different timescales, and knowing the difference matters whenever real value depends on a transaction staying put. This article explains what finality is, the two broad models for achieving it, and how Ethereum’s proof-of-stake system settles blocks.

Confirmed is not the same as final

A transaction is confirmed when it has been included in a block. But a recently added block is not necessarily permanent: under some conditions the network can end up building on a different version of history, in which case that block, and the transactions in it, can be dropped. Finality is the guarantee that this can no longer happen; a finalized transaction is settled for good. The gap between the two is why serious systems wait before treating a transaction as done. For the step-by-step path to confirmation, see how a transaction gets confirmed.

Why reversals are even possible

A blockchain is a chain of blocks that nodes must agree on. Occasionally two valid blocks appear at nearly the same time, or network delays cause nodes to briefly disagree about the latest block. The consensus rules resolve this by having the network converge on one version, which means the discarded blocks, and any transactions unique to them, are rolled back. This event is called a reorganization, or reorg. Reorgs are a normal, if uncommon, part of how chains settle disagreements at the tip. The crucial point is that they only affect recent, unfinalized blocks. To understand what is being reorganized, our explainer on what a blockchain actually is lays the groundwork.

It is worth being clear that a small, short reorg is not an attack or a bug; it is the system doing its job. When two valid blocks briefly compete, the network needs a rule to pick one and move on, and the losing block’s transactions return to the waiting pool to be included later. What separates a healthy reorg from a dangerous one is depth. Replacing the single most recent block is routine; replacing many blocks that most of the network has already built upon would require overwhelming resources and is what finality is designed to make effectively impossible.

Probabilistic finality

The first model is probabilistic finality. Here, a transaction is never declared absolutely irreversible; instead, reversing it becomes exponentially harder as more blocks are stacked on top of the one containing it. Each additional block deepens the transaction in the chain, and an attacker wanting to rewrite history would have to outpace the honest network’s block production from that depth onward, which quickly becomes impractical.

Proof-of-work networks such as Bitcoin work this way. This is the origin of the convention of waiting for a number of confirmations before considering a payment settled: more confirmations mean a deeper transaction and a smaller chance of reversal. It is a convention, not a hard protocol threshold, and the appropriate number depends on how much value is at stake. The trade-offs between this model and the alternative are discussed in proof of work vs proof of stake.

Deterministic and economic finality

The second model gives a firmer guarantee. Under deterministic finality, the protocol explicitly marks a specific block as final, after which it cannot be reverted. A closely related idea is economic finality: reversal is not mathematically impossible, but doing it would require destroying such an enormous amount of value that no rational actor would attempt it.

Ethereum’s proof-of-stake consensus uses this approach. It combines two mechanisms, Casper the Friendly Finality Gadget and the LMD-GHOST fork-choice rule, together referred to as Gasper. Rather than relying only on chain depth, validators actively vote to lock in the chain’s history, and the economic penalties for cheating are what make finalized blocks stick.

The distinction between the two models is not merely academic. Under probabilistic finality you are always making a judgement call: how many confirmations are enough for the value at stake? Under economic finality the protocol makes a firmer statement, drawing a line and saying that beyond it, reversal would demand the destruction of a specific, enormous fraction of the network’s security deposit. Neither model claims true mathematical impossibility, since a sufficiently catastrophic failure could in principle break either. The difference lies in how the guarantee is expressed and how the cost of breaking it is imposed.

How Ethereum finalizes a block

Ethereum divides time into slots of 12 seconds, grouped into epochs of 32 slots, about 6.4 minutes each. Only blocks at epoch boundaries, called checkpoints, can be justified and finalized. The process runs in two stages:

1. Justified: when validators representing at least two-thirds of the total staked ETH vote in favour of a checkpoint, it becomes justified. This is a strong signal but not yet final.
2. Finalized: when a later checkpoint is justified on top of a justified checkpoint, the earlier one becomes finalized, meaning it is committed to the canonical chain.

Because this requires votes across successive checkpoints, finality is reached over roughly two epochs, approximately 12.8 minutes under normal conditions. Until a block is finalized, it is still technically subject to a reorg; once finalized, it is not, short of a catastrophic consensus failure.

The cost of reversing finality

What makes finalized Ethereum blocks so durable is the price of undoing them. The documentation states that a finalized block “cannot be reverted unless there has been a critical consensus failure and an attacker has destroyed at least 1/3 of the total staked ether.” An attacker would need to control a large share of the stake and then have it destroyed by the protocol’s penalties for contradictory voting. This is economic finality in action: the guarantee rests not on impossibility but on a deliberately ruinous cost.

The mechanism behind that penalty is worth naming. Validators who vote in contradictory ways, the behaviour required to rewrite finalized history, are subject to slashing, in which a portion of their staked deposit is forcibly removed and they are ejected from the validator set. To finalize two conflicting versions of history at once, an attacker would have to get validators controlling at least a third of all stake to break the rules simultaneously and accept having that stake destroyed. The security therefore scales with how much value is staked: the more that is at stake, the more expensive an attack becomes. This ties the chain’s safety directly to its economics rather than to raw computing power.

Why finality matters in practice

Finality tells you when it is genuinely safe to act on a transaction. An exchange crediting a deposit, a bridge releasing funds on another chain, or a business shipping goods all need confidence that the transaction will not later vanish. Acting too early, on a confirmed-but-not-final transaction, is a real risk if a reorg follows.

This is especially acute for cross-chain systems. A bridge that assumes finality on one chain before it is truly settled can be exploited if that chain reorgs, which is one reason bridges are considered a weak point; see bridge risk: why cross-chain is the weak point. Different networks and layers finalize on different timescales, so a safe waiting period on one is not automatically safe on another. For the broader mechanics, our guide to how blockchains work connects finality to the rest of the settlement process.

How much to wait, and why it varies

There is no single universal answer to “how long until it is safe.” The right waiting period depends on the network’s finality model, the value at stake, and the counterparty’s own policy. On a probabilistic chain, a small payment might be treated as good after a handful of blocks, while a very large transfer warrants more. On a chain with explicit finality, the natural threshold is the point at which the block is finalized rather than merely included.

This is why different services quote different confirmation requirements for the same asset: each is making its own judgement about acceptable risk. It is also why a transaction that looks “done” in your wallet may still be pending from a recipient’s perspective. The wallet is reporting inclusion; the recipient may be waiting for finality. Understanding that these are two different milestones, not one, resolves a lot of apparent contradictions between what your wallet shows and when a deposit is actually credited. When you are the one relying on a transaction, the safe habit is to match your patience to the amount at stake rather than to the first green checkmark.

The takeaway

Finality is the difference between a transaction that is included and one that can never be taken back. Probabilistic finality, used by proof-of-work chains, strengthens with every block but never reaches certainty, which is why confirmations are counted. Deterministic and economic finality, used by Ethereum’s proof-of-stake system, lock in specific checkpoints after about two epochs and make reversal astronomically expensive. Whenever real value is on the line, the practical question is not “is it confirmed?” but “is it final?”

Sources

• Gasper — ethereum.org
• Proof-of-Stake (PoS) — ethereum.org
• Transactions — ethereum.org