What is Rollup?

How it works

A rollup runs its own execution environment (often EVM-compatible, so existing Ethereum contracts deploy unchanged). Users transact on the rollup at L2 prices. Periodically, the rollup operator submits the new state — what changed, who paid whom — back to Ethereum L1 in a compressed batch.

Two designs differ in how they prove validity:

• Optimistic rollups (Arbitrum, Optimism, Base) post the state transition along with the transaction data and assume it's correct. There's a "challenge period" (7 days on most chains) during which anyone can submit a fraud proof showing a bad transition; if accepted, the bad batch is reverted. In practice fraud has been rare.
• ZK rollups (zkSync Era, Polygon zkEVM, Linea, StarkNet) generate a zero-knowledge proof for each batch that mathematically verifies the state transition. Validity is checked instantly on L1; no challenge period needed. Proof generation is computationally expensive but improving fast.

Example

A typical L2 user flow:

1. Bridge 1 ETH from Ethereum L1 to Arbitrum (~$3 gas + ~10 min wait)
2. On Arbitrum, swap on Uniswap, lend on Aave, mint an NFT — each tx costs $0.05–0.50
3. Withdraw remaining ETH back to L1 (~7 day wait via canonical bridge, OR pay a third-party bridge fee for instant withdrawal)

The 7-day withdrawal delay on optimistic rollups exists because that's the challenge period — your withdrawal isn't safe until the challenge window closes. ZK rollups eliminate this; withdrawals can be ~hours instead of a week.

Why it matters

Rollups are how Ethereum scales without changing L1. Some practical implications:

• Most retail DeFi has moved to L2. Arbitrum, Optimism, and Base together processed more daily transactions than Ethereum L1 by 2024.
• Liquidity is fragmenting. A protocol like Aave deploys separately on every L2; each version has its own user base and liquidity. Cross-L2 unification is a hard, ongoing problem.
• EVM compatibility matters. Devs deploy the same Solidity to L2s with minimal changes — that's why "EVM-equivalent" is a marketing point for rollups.
• Centralization concerns. Most L2s have a single sequencer (transaction orderer) operated by the project team. Decentralizing sequencers is a 2024–2026 roadmap item across the major L2s.
• Data availability is the bottleneck. Rollups need to post transaction data somewhere. Posting to L1 (Ethereum) is expensive; alternative DA layers (Celestia, EigenDA) are cheaper but with weaker security assumptions.

For everyday users: pick the L2 with the most liquidity for the asset/protocol you're using (often Arbitrum or Base in 2024). Bridge canonical-route when stakes are high; use third-party bridges (Across, Stargate, Hop) for routine transfers.