Finality is the property that determines when a transaction can be considered truly settled — past the point where it could be reversed.
It matters more than most introductory explanations suggest. Every payment system has to answer this question. Traditional banking answers it with legal frameworks and central authority — if the bank says a wire cleared, it cleared. Blockchain systems answer it with cryptography and economic incentives instead, and the answer depends heavily on which chain you're using.
What "Final" Actually Means Varies by Protocol
The honest framing is that "finality" describes a spectrum, not a binary state.
On one end: probabilistic finality, where each additional block added after your transaction makes reversal exponentially less likely, but never mathematically impossible. Bitcoin works this way. After your transaction is included in a block, that block sits underneath every subsequent block. Reorganizing the chain to undo your transaction would require an attacker to redo all that proof-of-work faster than the honest network continues extending the chain. With enough confirmations, that attack becomes computationally infeasible — but it's economics, not cryptography, that provides the guarantee.
Six confirmations is the common rule of thumb for large Bitcoin transactions, which is roughly 60 minutes. This isn't a protocol rule — there's no flag that changes at confirmation six. It's a convention that emerged from analysis of what attack cost looks like after that many blocks. Some exchanges accept one or two confirmations for smaller amounts, which is fine; others require more for custody-level certainty.
On the other end: deterministic finality, where the protocol guarantees irreversibility at a specific point. Byzantine Fault Tolerant (BFT) consensus systems like Cosmos use Tendermint consensus, which works this way. Once a block is committed, it cannot be reverted. The validators have signed it; reversing it would require colluding validators to provably misbehave. There's no "waiting for more confirmations" — the first confirmation is final.
Ethereum Sits in Between: Economic Finality
Ethereum post-Merge introduces a third category worth understanding separately.
Time on Ethereum is divided into slots (12 seconds each) and epochs (32 slots, about 6.4 minutes). At each epoch boundary, there's a checkpoint. Validators vote on these checkpoints using a component called Casper FFG — the Friendly Finality Gadget — layered on top of Ethereum's main fork-choice rule.
When two-thirds of all staked ETH votes for a checkpoint, that checkpoint becomes justified. When two consecutive checkpoints are justified, the earlier one becomes finalized. This typically takes two epochs — roughly 13 minutes from when your transaction was included.
What makes this "economic" rather than "deterministic" is what reverting would cost. Undoing a finalized checkpoint would require validators representing more than one-third of all staked ETH to get slashed — they'd lose their entire stake. With roughly 30 million ETH staked as of mid-2026, the cost of reverting a finalized checkpoint is on the order of tens of billions of dollars in ETH that would be destroyed.
This isn't just theoretical. Ethereum's slashing conditions are automatic: any node on the network can detect conflicting votes and submit evidence, triggering the slash. An attacker can't back out quietly.
Where the Tradeoffs Live
BFT systems achieve finality fastest, but they require validators to communicate in multiple rounds per block. More validators means more messages, which means slower rounds. Ethereum can't use a pure BFT approach at its current scale — the validator count (well over 1 million active validators as of mid-2026) makes the communication overhead prohibitive.
Bitcoin's probabilistic approach doesn't have this constraint; validators don't coordinate directly. The trade-off is accepting that "final" means "extremely unlikely to reverse" rather than "impossible to reverse."
This is sometimes framed as a trilemma tradeoff: achieving fast finality while maintaining a large, decentralized validator set is genuinely hard. The systems that achieve fast deterministic finality tend to have smaller, more tightly coordinated validator sets.
What's Changing: Single Slot Finality
Ethereum's roadmap includes Single Slot Finality (SSF), which would compress the ~13-minute finality window down to a single 12-second slot. Instead of waiting for two epochs of checkpoint votes, finality would happen within the slot where the block is proposed.
This is technically challenging at Ethereum's scale. The main obstacle is validator set size — the protocol would need to aggregate votes from a very large number of validators within 12 seconds. Current research is focused on making this feasible without reducing the validator count, which would be a centralization regression.
SSF is still in the research phase. No production timeline has been committed.
What Would Confirm This Direction
A successful SSF specification that handles vote aggregation at Ethereum's scale, followed by mainnet deployment without incident, would confirm the roadmap is executable. For the current economic finality model, sustained operation without a finalization failure or a successful slashing-based reorg attempt would confirm the model's robustness.
What Would Break or Invalidate It
A finalized Ethereum checkpoint being successfully reversed would invalidate the economic finality model — it would require a coordinated attack burning more than $10B in staked ETH. A successful multi-block reorg on Bitcoin at the six-confirmation level, actually executed on mainnet, would invalidate the probabilistic finality assumptions most exchanges and payment processors rely on. For SSF: if the validator set would need to be significantly reduced to make SSF feasible, that would represent a security regression that may block the upgrade entirely.
Timing
Now: Ethereum finality takes ~13 minutes; Bitcoin finality is probabilistic with six confirmations (~60 minutes) as the standard for large transactions; BFT chains like Cosmos finalize per block. None of this is changing imminently.
Next: SSF research is ongoing. No deployment timeline.
Later: Cross-chain finality — guaranteeing settlements across two different chains are atomic and simultaneous — is an unsolved problem at the protocol layer.
Boundary
This covers the mechanism of finality — when and how blockchains decide a transaction is settled. It doesn't address finality in the context of optimistic rollup fraud proof windows, cross-chain bridge security models, or how individual applications should set their own confirmation thresholds. Those depend on risk tolerance and context outside this scope.
A transaction past finality is settled by the chain's rules. Whether that settlement transfers to your specific risk framework is a separate question.
