Layer 2s Compared: Fees, Speed, and Real Trade-offs
Rollups are cheaper because they batch. What differs between them is how withdrawals work and who can stop them.
A layer 2 processes transactions off the main chain and periodically posts data back to it. Users split the cost of that settlement, which is where the saving comes from.
The differences between them are less about fees, which are all low, and more about how you get your money out.
The two designs
Optimistic rollups assume transactions are valid and allow a challenge period during which anyone can prove fraud. That period is typically seven days, and it is why a standard withdrawal to mainnet takes a week.
Zero-knowledge rollups post a cryptographic proof that the transactions were executed correctly. Nothing needs to be assumed, so withdrawals can settle in hours rather than days once the proof is posted.
What actually differs
| Property | Optimistic | Zero-knowledge |
|---|---|---|
| Standard withdrawal to mainnet | About 7 days | Hours |
| Transaction cost | Very low | Very low |
| Proving overhead | None | Computational cost, paid in fees |
| Compatibility with Ethereum tooling | Generally complete | Varies by implementation |
| Maturity | Longer track record | Newer, improving quickly |
The seven-day withdrawal is the single most practically important difference, and it is routinely bypassed by third-party bridges that front you the funds on mainnet immediately for a fee. That is a trade: speed in exchange for trusting the bridge operator.
The shared risks nobody advertises
Sequencer centralisation. On most rollups today, a single operator orders transactions. If it stops, you cannot transact normally. All major rollups have a mechanism for forcing transactions through mainnet directly, which is slow and expensive and exists precisely for that scenario.
Upgrade keys. Most rollup contracts can be upgraded by a multi-signature wallet held by the development team. This is a real trust assumption and it is documented publicly for each network. Whether the keys are held by a small team or by a broader set with a timelock is worth checking.
Bridge risk. Bridges have historically been among the most exploited components in crypto. The canonical bridge operated by the rollup itself is generally safer than third-party fast bridges, and slower.
What to check before using one
- Who runs the sequencer, and what happens if it stops
- Who can upgrade the contracts, and whether there is a timelock
- Whether the canonical bridge is being used or a third party
- Actual current fees for the operation you intend, not the headline figure
The first three are published. Several independent sites track the trust assumptions of each rollup in a standard format, which is the fastest way to compare them honestly.
The practical position
For ordinary use, the differences matter less than the decision to use a rollup at all. Moving from mainnet to any established layer 2 cuts costs by a factor that makes the choice between rollups a secondary question.
The withdrawal timeline is the thing to plan around. If funds may be needed on mainnet quickly, a zero-knowledge rollup or a plan that accounts for seven days is the relevant consideration.
For moving between the chain and ordinary currency, going through an exchange that supports withdrawals directly to the rollup, such as an exchange with a published withdrawal schedule, avoids the bridge entirely and is usually the cheapest route in both directions.
Fees move with network demand. Any figure here reflects the date shown above and should be re-checked before you act on it. Where a measurement was taken by hand, the article says so.
Related reports
- Rollup Fee Composition: Where the Fraction of a Cent Goeslayer 2 ecosystem
- What Rollups Actually Pay For, and Why Their Fees Felllayer 2 ecosystem
- Gas Fees Across Chains: A Monthly Comparisonlayer 2 ecosystem