If you have swapped a token on Base, bridged into Arbitrum, or minted something on a chain that felt oddly cheap, you have already used a rollup. Most people never learn the name for it. Layer 2 rollups are the quiet plumbing that lets Ethereum serve millions of users without asking each of them to pay for a slot in a block that only fits a few hundred transactions.
The idea is older than the hype around it. Ethereum can process a limited number of transactions per second on its base layer, and when demand spikes, the price of that space climbs fast. Rollups exist to move the heavy work somewhere cheaper while keeping Ethereum as the court of final record. This piece walks through what that actually means, how the two main designs differ, and why fees fell off a cliff in 2024.
Ethereum's block space is small on purpose
Every node in the network re-runs every transaction to check it. That redundancy is what makes Ethereum hard to cheat, and it is also what caps throughput. There is only so much a global set of home computers can verify per block. When a popular mint or a market panic fills those blocks, users bid against each other and gas fees spike into the tens of dollars.
The base layer can grow that capacity a little. A recent upgrade path lifted Ethereum's gas limit, which is roughly how much computation fits in a block, and even that modest change reshuffled long-standing assumptions about the network, as our look at the Glamsterdam gas rule showed. But raising the limit endlessly would push out the smaller machines that keep the chain decentralized. So the network took a different route. Instead of doing more work on Ethereum, do it somewhere else and post the receipts back.
How a rollup actually works
A rollup runs its own chain. Transactions happen there, fast and cheap, because a single sequencer orders them rather than a whole planet of validators. Then the rollup takes a large batch of those transactions, compresses the data, and writes it to Ethereum along with a claim about the new state of its ledger.
That last part is the whole trick. The rollup does not ask you to trust its sequencer. It posts enough information to Ethereum that anyone can reconstruct what happened and, crucially, that Ethereum itself can enforce the outcome if the operator lies. Hundreds of trades collapse into one L1 footprint, and the cost of that footprint is split across everyone in the batch. Security is inherited from Ethereum. Speed and price come from doing the busywork off chain.
The split between the two families of rollups comes down to a single question. How does Ethereum know the batch is honest?
Fraud proofs and the seven day wait
An optimistic rollup takes the batch at its word. It posts the data and says, in effect, this is valid unless someone proves otherwise. That someone is a watcher running a fraud proof. If the batch contains an invalid transaction, a challenger submits evidence, Ethereum re-checks the disputed step, and a bad batch gets thrown out.
The catch lives in the waiting. Because the network has to leave time for an honest challenger to notice fraud and respond, withdrawals back to Ethereum sit behind a challenge window. On the two largest optimistic rollups, Arbitrum and Optimism, that window runs about seven days. Move funds within the rollup and it feels instant. Pull them out to the base layer and you wait a week, unless you pay a third party to front the money. Arbitrum, Optimism, and Base all use this model.
ZK rollups prove every batch with math
A zero-knowledge rollup flips the assumption. Nothing is accepted until it is proven. For each batch, the rollup generates a validity proof, a compact piece of cryptography that mathematically demonstrates the new state follows from the old one under the rules. Ethereum verifies the proof and updates. No batch, no proof, no update.
Because the proof settles correctness on the spot, there is no week-long challenge window. Exits can finalize in hours rather than days. The cost sits elsewhere. Generating those proofs is computationally demanding, and building a ZK system that runs general smart contracts took years longer than the optimistic approach. zkSync, Starknet, Linea, and Scroll live in this camp.
A side by side comparison
| Property | Optimistic rollup | ZK rollup |
|---|---|---|
| Core assumption | Valid unless challenged | Invalid until proven |
| Proof type | Fraud proof, submitted only on dispute | Validity proof, posted with every batch |
| Withdrawal to Ethereum | About 7 days | Hours |
| Main cost | Delay and watcher assumptions | Heavy proof computation |
| Examples | Arbitrum, Optimism, Base | zkSync, Starknet, Linea, Scroll |
Blobs made rollup data cheap in 2024
For a while, rollups had a dirty secret. Posting their batch data to Ethereum was expensive, because they were renting the same permanent storage as everything else on the base layer. That cost got passed to users. A cheap rollup transaction still carried a not-so-cheap slice of L1 data fees.
EIP-4844, which shipped with the Dencun upgrade on March 13, 2024, changed the math. It introduced a new transaction type carrying blobs, chunks of data around 128 kilobytes each that Ethereum stores temporarily and prunes after roughly 18 days. Rollups do not need their data kept forever. They need it available long enough for anyone to verify a batch. Blobs give them exactly that, in a separate fee market from regular gas.
The effect was immediate. One analysis measured an 81% drop in the calldata that optimistic rollups had to pay for. Median fees on Arbitrum, Optimism, Base, and zkSync fell from tens of cents to fractions of a cent almost overnight. A later modelling estimate tied to the Fusaka upgrade in December 2025 suggested rollup costs could fall a further 40 to 60% as blob capacity ramps, though that figure comes from a single projection and should be read as a forecast, not a settled result.
The tradeoffs rollups introduce
Cheaper and faster came with a cost that is easy to miss. Ethereum used to be one place. Now it is a hub with dozens of rollups orbiting it, each with its own liquidity, its own bridges, and its own quirks. Moving between them is not free and not always simple, and the fragmentation shows up in the data. Two respected trackers once measured total value on Ethereum's rollups and landed billions of dollars apart, a gap that says a lot about how young this layer still is.
There is also the sequencer question. Most rollups today run a single operator that orders transactions, which is efficient and also a point of control. If it goes offline or censors, users depend on escape hatches that vary in quality from chain to chain. The industry is working toward shared and decentralized sequencing, but the fragmented L2 economy is a real and current tradeoff, not a solved problem. Data availability is the other quiet dependency. A rollup is only as safe as its promise that anyone can fetch the batch data and rebuild the ledger, so where and how that data is published matters as much as the proofs sitting on top of it.
FAQ on Layer 2 rollups
Is a rollup the same as a sidechain?
No. A sidechain runs its own security, so if it fails, your funds are at risk on that chain alone. A rollup posts its data and proofs to Ethereum, which means Ethereum can enforce correct behavior. That inherited security is the line between the two.
Are my funds as safe on a rollup as on Ethereum?
Close, but read the fine print. The strongest rollups let anyone reconstruct state from Ethereum and force an exit even if the operator disappears. Younger ones still carry admin keys and upgrade powers that could, in theory, be misused. Trackers like L2Beat grade these differences openly.
Why does an optimistic rollup make me wait to withdraw?
The delay is the security. The challenge window gives honest watchers time to catch and prove a fraudulent batch before funds leave the system. Remove the wait and you remove the window in which fraud can be stopped.
Do ZK rollups make Ethereum obsolete?
They lean on it harder, not less. Every validity proof is verified on Ethereum, and the batch data lives there. Rollups scale Ethereum by borrowing its trust, which keeps the base layer central to the whole arrangement.
Rollups turned Ethereum from a network that priced out ordinary users during busy weeks into one where a swap can cost less than a cent. The design is not finished. Exit times, sequencer control, and the mess of moving value between chains are all live work. But the core bet has held up in practice, and for most people transacting on Ethereum today, a rollup is doing the work whether they know it or not.
