Layer 2 Crypto Explained: ZK vs Optimistic Rollups

A layer 2 (L2) in crypto is a separate network that executes transactions off a main blockchain, the layer 1 (L1), and then posts the results back to it so the L1 can secure them. The best-known layer 2s are rollups built on Ethereum: they bundle thousands of transactions, run them cheaply, and publish compressed data plus a proof or a claim to Ethereum. The two families are optimistic rollups, which assume batches are valid unless someone proves fraud, and ZK rollups, which prove every batch with a validity proof.

This guide explains how rollups work, the difference between optimistic and ZK rollups, what blobs changed, how to read L2BEAT's stages, which L2s hold the most value right now, and the risks that come with sequencers and bridges. It also covers chains such as Sui and MultiversX that scale on the base layer instead.

L1 vs L2: what's the difference?

A layer 1 is a blockchain that settles its own transactions and secures itself with its own validators: Ethereum, Bitcoin, Sui and MultiversX are all L1s. A layer 2 runs on top of an L1 and borrows its security for the things that matter most: the ordering of the data, the validity of the state, and your ability to withdraw.

The reason L2s exist is cost. Every Ethereum node re-executes every transaction, so block space is scarce and fees rise when demand spikes. Our guide to gas fees explains that auction. A rollup moves execution elsewhere and only asks Ethereum to store data and check a proof or a dispute, which costs far less per transaction.

"Inherited security" has limits. A rollup is only as safe as its smart contracts, its proof system and whoever can upgrade those contracts, which is why L2BEAT's stages, covered below, exist.

How rollups work

Every rollup follows the same basic loop:

1. Users send transactions to the sequencer. The sequencer orders them and gives you a fast "soft" confirmation before anything reaches Ethereum.
2. The sequencer batches them. Thousands of transactions are compressed into one batch.
3. The batch data goes to Ethereum. According to ethereum.org, rollups post transaction data to mainnet as calldata or in blobs, so anyone can rebuild the rollup's state from Ethereum alone.
4. A new state root is proposed. A smart contract on Ethereum records the claimed result of the batch.
5. The result is checked. Optimistic rollups leave a window for anyone to dispute it; ZK rollups submit a cryptographic proof that the contract verifies.

Step 3 is what makes something a rollup. If the data lives on a separate committee or network instead, L2BEAT classifies it as a validium (ZK) or an optimium (optimistic), which trades some security for lower cost.

Optimistic rollups vs ZK rollups

Optimistic rollups

An optimistic rollup assumes each batch is correct. Anyone watching the chain can challenge a batch with a fraud proof; if the challenge succeeds, the bad state is reverted and the dishonest party's bond is slashed. To give watchers time, withdrawals to Ethereum wait out a challenge period. Ethereum.org says users "must wait for the one-week challenge period to expire before withdrawing funds back to Ethereum" (source). Third-party bridges and liquidity providers will front you the funds sooner, for a fee.

Base, Arbitrum One and OP Mainnet are all optimistic rollups. Their big advantage is that they run the Ethereum Virtual Machine almost unchanged, so contracts and tools port across easily.

ZK rollups

A ZK rollup attaches a validity proof (a SNARK or a STARK) to each batch, and the Ethereum contract only accepts the new state if the proof checks out. Per ethereum.org, there are no challenge delays when moving funds from a ZK rollup to Ethereum once the proof is verified. The costs are elsewhere: generating proofs is computationally heavy, verifying one on mainnet costs roughly 500,000 gas, and proving general EVM execution (a "zkEVM") is harder than proving simple transfers.

Starknet, ZKsync Era, Linea and Scroll are examples, all classed as ZK rollups by L2BEAT. The zero-knowledge part is about succinct proofs, not privacy: most ZK rollups publish their transaction data openly.

Data availability and blobs (EIP-4844)

For most rollups, the biggest cost used to be publishing data to Ethereum. That changed with the Dencun upgrade, which went live on 13 March 2024 at epoch 269568 (Ethereum Foundation). Dencun introduced proto-danksharding, EIP-4844, which added "blobs": chunks of data with their own fee market that are pruned after roughly 18 days. Rollups only need the data to be available long enough for anyone to check or rebuild the state, so temporary storage is enough and much cheaper than permanent calldata.

Blob capacity has been raised since:

• Pectra (7 May 2025): EIP-7691, activated in the Pectra upgrade, raised the per-block blob target and maximum from 3 and 6 to 6 and 9.
• Fusaka (3 December 2025): introduced PeerDAS, which lets nodes check that blob data is available by sampling pieces of it instead of downloading everything (Ethereum Foundation).
• Blob-parameter-only forks: BPO1 on 9 December 2025 raised the target and maximum to 10 and 15; BPO2 on 7 January 2026 raised them to 14 and 21 (same source).

More blob space generally means cheaper L2 transactions, though fees on any given day depend on demand.

L2BEAT stages: how decentralised is your L2?

Many rollups launched with "training wheels": an operator or multisig that can upgrade contracts instantly or override proofs. L2BEAT's stages framework grades how far each project has removed them:

• Stage 0: the project posts state roots and data to the L1, publishes software that can rebuild its state, and has a proper proof system, but operators can still override it.
• Stage 1: apart from bugs, blocking or forcing invalid withdrawals requires compromising at least 75% of a Security Council, and upgrades from outside that council must give users at least a 7-day exit window.
• Stage 2: the proof system is open to anyone, users get at least 30 days to exit unwanted upgrades, and the Security Council can only step in for bugs that can be proven onchain.

As of 17 October 2026, L2BEAT had assigned a stage to 67 layer-2 projects: 57 were Stage 0, six were Stage 1 and four were Stage 2. The four Stage 2 projects held almost no value. Check the stage before you park large sums on any L2.

Top layer 2s by value secured

L2BEAT measures total value secured (TVS): assets bridged in through the canonical bridge, assets bridged through external bridges, and tokens minted natively on the L2. The largest Ethereum rollups as of 17 October 2026, per L2BEAT:

1. Base (optimistic, OP Stack, Stage 1): about $15.4bn.
2. Arbitrum One (optimistic, Stage 1): about $10.4bn.
3. OP Mainnet (optimistic, OP Stack, Stage 1): about $2.0bn.

The largest ZK rollups were Mantle (about $1.3bn) and Lighter (about $1.1bn), both Stage 0; Starknet, at about $0.57bn, was the biggest ZK rollup at Stage 1. L2BEAT also lists projects that aren't rollups by its definition, such as Polygon PoS, under separate categories, so headline rankings depend on which list you read.

The mood around L2s has also shifted. On 3 February 2026 Vitalik Buterin wrote on X that "the original vision of L2s and their role in Ethereum no longer makes sense, and we need a new path", citing slow progress on decentralisation and a base layer that now scales and keeps fees low (The Defiant). He suggested L2s focus on things beyond raw scaling, such as privacy and app-specific features. Our blockchain trilemma guide covers the trade-offs behind that debate.

Layer 2 risks: sequencers, upgrades and bridges

Sequencer risk

Major L2s such as Base rely on a sequencer run by a single operator. It can reorder transactions (a source of MEV), delay them, or stop. On 5 August 2025 Base halted block production for 33 minutes after a sequencer handoff failed, which paused deposits and withdrawals too (The Block). Good designs let users force a transaction through the L1 if the sequencer censors or fails: L2BEAT lists "self sequence" for Base, Arbitrum One and OP Mainnet, with up to a 12-hour delay on Base.

Upgrade risk

Whoever controls the rollup's contracts on Ethereum can, in a Stage 0 system, change the rules, including how withdrawals work. This is the risk the stages framework measures.

Bridge risk

The canonical bridge is the rollup's own contract on Ethereum, and it carries the rollup's security. Third-party bridges that move assets between L2s or out quickly are separate systems with their own validators or liquidity, and bridges have been frequent targets of large hacks. Our crypto bridge guide walks through the types.

Scaling without a layer 2: Sui and MultiversX

Rollups are Ethereum's answer to scaling. Other L1s scale on the base layer, so there is no separate network, sequencer or bridge to trust.

• MultiversX uses adaptive state sharding: the chain is split into execution shards plus a coordinating metachain. The live network configuration showed 3 shards, 400 nodes per shard and a 600 ms round duration when we checked on 17 October 2026 (MultiversX gateway). See what is sharding.
• Sui orders transactions with Mysticeti, a DAG-based consensus that the Sui docs say commits in about 0.5 seconds on average in benchmarks, and executes transactions that don't touch the same objects in parallel. See what is Mysticeti.

Neither approach is free: sharding makes cross-shard calls asynchronous, and high-throughput L1s need more powerful validator hardware.

Where JewelSwap fits

JewelSwap runs on L1s rather than L2s: liquid staking (JWLSUI, JWLEGLD, JWLXRD) on Sui, MultiversX and Radix, plus Sui farms built on Scallop and Cetus. Using it means no rollup withdrawal window, but the usual smart contract and market risks still apply.

Frequently asked questions

What is a layer 2 in crypto?

A layer 2 is a network that executes transactions off a main blockchain and posts the data and results back to it, so the main chain secures them. The best-known layer 2s are rollups on Ethereum, such as Base, Arbitrum One and OP Mainnet.

What is the difference between a ZK rollup and an optimistic rollup?

An optimistic rollup assumes batches are valid and lets anyone challenge them with a fraud proof during a challenge period of about a week. A ZK rollup submits a validity proof with every batch, so Ethereum checks correctness up front and withdrawals don't wait for a challenge window.

Why do optimistic rollup withdrawals take 7 days?

The delay gives anyone watching the chain time to submit a fraud proof against an invalid batch before funds leave the rollup. Third-party bridges can pay you out sooner for a fee, but you then rely on that bridge.

What did EIP-4844 do for layer 2s?

EIP-4844, activated in the Dencun upgrade on 13 March 2024, added blobs: temporary data space with its own fee market, pruned after about 18 days. Rollups post their data in blobs instead of permanent calldata, which made L2 transactions much cheaper.

Which layer 2 is the biggest?

As of 17 October 2026, Base was the largest Ethereum layer 2 by total value secured on L2BEAT, at about $15.4bn, followed by Arbitrum One at about $10.4bn and OP Mainnet at about $2.0bn.

Are layer 2s safe?

They are only as safe as their proof system, their upgrade controls and their sequencer setup. Check the project's L2BEAT stage: most were still Stage 0 in October 2026, meaning operators could override the rules.

Keep reading

• The blockchain trilemma explained
• What is a crypto bridge?
• What are gas fees?
• What is sharding?
• What is Mysticeti?

This article is educational and isn't financial advice. Figures are as of 17 October 2026 from L2BEAT, the Ethereum Foundation, ethereum.org and the MultiversX and Sui documentation, and they change often; sources are linked throughout.