Layer 2s: Rollups and Scaling

A Layer 2 (L2) is a separate network that executes transactions cheaply and quickly, then posts compressed records of them back to a Layer 1 blockchain — usually Ethereum — which acts as the final judge of what happened. The result is fees that are often cents instead of dollars, while security still leans on the underlying chain rather than on a brand-new validator set. The dominant L2 design today is the rollup, and the two main families — optimistic and zero-knowledge (ZK) — differ mainly in how they prove to Ethereum that their transactions were valid.

Why Layer 2s Exist

Ethereum deliberately keeps the requirements for running a node low so that many independent parties can verify the chain. The price of that decentralization is limited capacity: on the order of tens of transactions per second on the base layer. When demand spikes, users bid against each other for block space, and gas fees climb. During peak periods in past cycles, a simple token swap on Ethereum mainnet could cost $50-100 or more. That's tolerable for settling a $100,000 transfer; it's absurd for a $30 one.

The scaling approach Ethereum settled on is often called a rollup-centric roadmap: instead of making the base chain enormously bigger (which raises hardware requirements and centralizes validation), push execution off-chain to L2s and use the L1 for what it's best at — being a hard-to-corrupt settlement and data layer.

A rollup batches hundreds or thousands of user transactions, executes them on its own network, and posts two things to Ethereum: the compressed transaction data and a claim about the resulting state (everyone's new balances). Because thousands of transactions share the cost of one L1 posting, each user's share of that cost is small. Ethereum's Dencun upgrade in March 2024 added "blob" data space specifically for rollups, which cut L2 costs dramatically — common transactions on major rollups routinely cost a few cents or less afterward.

The key security property: because the transaction data lives on Ethereum, anyone can reconstruct the rollup's state and check the claims. The L2's operators can't quietly invent balances, and even if they vanish, users can in principle exit using the data on L1.

Optimistic Rollups: Innocent Until Proven Guilty

Optimistic rollups (Arbitrum and OP Mainnet/Optimism are the best-known) post state claims to Ethereum without proof, assuming they're honest. The safeguard is a challenge window — typically about seven days — during which anyone watching can submit a fraud proof demonstrating that a posted state is wrong. If a challenge succeeds, the bad state is rolled back and the dishonest party is penalized.

This design is simple and mature, and it was the first to support the full Ethereum development environment, which is why so many apps deployed there early.

The trade-offs:

Slow native withdrawals. Moving funds from the L2 back to Ethereum through the official bridge takes roughly the length of the challenge window — about a week — because Ethereum must wait to be sure no fraud proof arrives. Third-party "fast bridges" front you the money immediately for a fee and collect the slow withdrawal themselves.
Someone must watch. The security model assumes at least one honest, capable party is monitoring and ready to challenge fraud. In practice this has held, but it is an assumption, not a proof.

ZK Rollups: Guilty Until Proven Valid

ZK rollups (such as zkSync Era, Starknet, Scroll, and Linea) take the opposite approach: every batch comes with a validity proof — a compact cryptographic proof that all transactions in the batch followed the rules. Ethereum verifies the proof before accepting the state. There is nothing to challenge, because invalid states can't be accepted in the first place.

Consequences:

Faster finality to L1. Once the proof is verified on Ethereum, the state is final — no week-long window. Withdrawals are limited mainly by how often proofs are generated and posted, typically hours rather than days.
Heavier machinery. Generating validity proofs is computationally expensive, and building proof systems that handle Ethereum-style smart contracts is genuinely hard. This tech matured later than optimistic rollups, and proving costs are part of the fee structure.

The names can mislead: the "zero-knowledge" part (hiding information) mostly isn't the point here — these systems use the proof machinery for validity, not privacy.

What Actually Changes for You as a User

Day to day, using an L2 feels like using Ethereum with smaller numbers.

Getting there. You either bridge assets from Ethereum (via the official bridge or a fast bridge) or, often more cheaply, withdraw directly to the L2 from an exchange that supports it — many major exchanges let you withdraw ETH or USDC straight to Arbitrum, Optimism, or Base.
Same wallet, different network. Your existing address works on most L2s; you add the network to your wallet and pay gas in ETH (on most rollups). A swap that might cost $5-20 on mainnet in busy conditions typically costs a few cents to a few tens of cents on a major L2.
Same asset, different ledger. ETH on Arbitrum is not the same ledger entry as ETH on mainnet. Sending tokens to the right address on the wrong network is a classic mistake — always confirm which network the recipient (especially an exchange deposit address) actually supports.
Getting out. Know the exit path before you enter: about a week via an optimistic rollup's native bridge, faster via ZK rollups or fast-bridge services, or instantly by depositing to an exchange that supports that network.

The Honest Caveats

L2s inherit much of Ethereum's security, but "much" is doing work in that sentence.

Centralized sequencers. On most major rollups today, a single operator (the sequencer) orders transactions. It generally cannot steal funds or forge state, but it can go down — briefly halting the network — or theoretically censor transactions. Escape hatches let users force transactions through L1, but they're clunky. Decentralizing sequencers is on every major team's roadmap and has been for years.
Upgrade keys. Many rollups can still be upgraded by a multisig or security council. A malicious or compromised upgrade is a real, if remote, risk. Public trackers grade rollups on how far they've progressed toward full trust-minimization ("training wheels" stages), and most major rollups are still partway there.
Not everything called an L2 is a rollup. Some chains market themselves as L2s while keeping transaction data off Ethereum (validiums, optimiums) or being independent sidechains with their own validators. These are legitimate designs but weaker security inheritances — if the data isn't on the L1, your exit guarantee depends on whoever holds that data.
Fragmentation. Liquidity and apps are spread across many L2s, so you may find yourself bridging between them, with all the usual bridge considerations.

Key Takeaways

L2 rollups execute transactions cheaply off-chain and post compressed data to Ethereum, which anchors their security; users get fees of cents instead of dollars.
Optimistic rollups assume validity and allow about a week for fraud proofs, which is why native withdrawals are slow; ZK rollups prove validity up front and finalize faster.
For users, L2s feel like Ethereum with lower fees — but ETH on an L2 is a separate ledger, and sending to the wrong network is an expensive, common mistake.
Most rollups still run centralized sequencers and hold upgrade keys, so their trust-minimization is real but incomplete; check a rollup's maturity before parking large sums.
Always know your exit route — native bridge, fast bridge, or exchange — and its cost and delay before you move funds in.