What Are ZK Rollups?

ZK rollups are layer 2 networks that execute transactions off the main chain, then submit a zero-knowledge validity proof to Ethereum alongside each batch. The proof is a small piece of cryptographic data showing that every transaction in the batch followed the rules, so Ethereum can verify the result without re-executing anything.

Because correctness is proven before a batch settles, ZK rollups reach hard finality on Ethereum quickly. Withdrawals clear once the proof is verified, typically within minutes to a few hours, so users never sit through a dispute window.

The historical trade-off was cost and complexity, since covering every operation of the Ethereum Virtual Machine took years of specialized engineering. That barrier has largely collapsed, and general purpose proving systems such as Succinct's SP1, RISC Zero, and OpenVM can now prove ordinary Ethereum execution written in standard code, so most new chains launching today choose validity proofs from the start.

What Are Optimistic Rollups?

Optimistic rollups publish transaction results to Ethereum and assume they are honest. Anyone running a verifier can challenge a batch during a dispute window, usually seven days, by submitting a fraud proof that forces Ethereum to re-check the contested computation and revert it if it fails.

The design is simpler to build and cheap to run because no proof is generated in normal operation. That head start explains why the three largest general purpose layer 2 networks, Base, Arbitrum One, and OP Mainnet, all use optimistic architectures and hold most of the ecosystem's value.

The price is the wait. A withdrawal through the native bridge takes the full challenge period to finalize, so users either hold on for a week or pay a third party bridge to front them liquidity. The model also assumes at least one honest party is watching the chain and ready to challenge fraud in time, an assumption that cryptographic proofs remove.

ZK Rollups vs Optimistic Rollups: Key Differences

Both designs inherit Ethereum's security by posting their data to mainnet, and both became dramatically cheaper after EIP-4844 introduced blob storage. The meaningful differences sit in how each one validates state.

• State validation: ZK rollups prove every batch is correct before it settles. Optimistic rollups settle first and rely on fraud proofs to catch invalid batches afterwards.
• Withdrawal finality: ZK withdrawals clear once the validity proof is verified on Ethereum, usually within hours. Optimistic withdrawals through the canonical bridge wait out the seven day challenge window.
• Security assumptions: ZK security rests on the soundness of the proof system and the correctness of the verifier contract. Optimistic security additionally requires an honest watcher to spot and challenge fraud before the window closes.
• Cost structure: Optimistic chains skip proving costs in normal operation. ZK chains pay for every proof, though prover expenses have fallen roughly 45-fold as hardware and proving systems matured.
• EVM compatibility: Optimistic rollups reuse Ethereum clients almost unchanged. Modern zkEVMs have closed most of the gap, so compatibility no longer decides the choice.
• Ecosystem depth: Optimistic chains hold the deepest liquidity and application coverage. ZK designs lead in appchains that need fast settlement, such as the perpetuals exchange Lighter.

Which Rollup Type Is Safer?

On paper, validity proofs give the stronger guarantee. A ZK rollup cannot finalize an invalid state as long as the math and the verifier contract are sound, whereas an optimistic rollup depends on someone noticing fraud within the window. In practice, we first weigh how much control the operator still holds.

L2Beat grades every rollup on a staging system. Stage 1 means the proof system runs permissionlessly, users can exit to Ethereum without the operator's cooperation, and upgrades face a delay that gives users time to leave. Stage 0 means a multisig held by the team can still override the system.

As of early August, Base, Arbitrum One, OP Mainnet, Ink, Unichain, and Starknet sit at Stage 1, while Linea, ZKsync Era, and Scroll remain at Stage 0 despite their stronger proof architecture.

Young proving systems carry risks of their own. Several STARK based provers rely on mathematical conjectures that researchers have recently pressured, which is why Ethereum's developers now require 128-bit provable security before zkEVMs touch consensus. Until a ZK chain pairs mature proofs with Stage 1 decentralization, the architectural advantage on its own settles very little.

The Line Between the Two Is Blurring

To see where scaling is heading, watch the optimistic camp. Proving has become a commodity service, and tooling such as OP Succinct lets any OP Stack chain swap fraud proofs for validity proofs without rebuilding its infrastructure.

Base, the largest layer 2 by value, is partway through that move. Its Azul upgrade adds SP1 zero-knowledge proofs alongside trusted hardware attestations in a multi-proof design, cutting withdrawal finality from seven days to one when both proofs agree, and the team describes it as a staging post toward full ZK proving.

Mantle and Taiko have gone further and now settle with validity proofs despite launching as optimistic chains, and the Bitcoin focused rollup BOB runs a hybrid system that resolves fraud disputes with a single ZK proof instead of a lengthy challenge game.

We have found no example of a validity proven chain converting back to fraud proofs, so the old labels increasingly describe where a chain started rather than where it is going.

Leading Examples of Each

The projects below anchor each camp today. For deeper coverage of the ZK side, see our ranking of the best ZK rollup projects.

ZK Rollups

• Starknet: Runs its own Cairo language and the Stwo prover, and is the first sizeable ZK chain to reach Stage 1 decentralization.
• Linea: Built by Consensys, the company behind MetaMask, with a design closely aligned to Ethereum's own economics.
• ZKsync Era: The flagship of the ZKsync stack, whose Boojum prover also powers chains such as Abstract.
• Scroll: Targets bytecode level Ethereum compatibility and proves execution with the OpenVM system.
• Lighter: An application specific ZK rollup for perpetuals trading, and the most active rollup of any kind by user operations.

Optimistic Rollups

• Base: Coinbase's chain and the largest rollup by value secured, now adding ZK proofs through the Azul upgrade.
• Arbitrum One: The long-standing DeFi heavyweight, secured by the permissionless BoLD challenge protocol.
• OP Mainnet: The original OP Stack chain, whose codebase also underpins Base, Ink, Unichain, and the wider Superchain.
• Ink: Kraken's DeFi focused network, already rated Stage 1 within a year of launch.
• Unichain: Uniswap's trading oriented chain, built to concentrate swap liquidity on its own block space.

Rollup Statistics and Adoption

L2Beat data from early August shows how the two camps compare in scale.

• Total value secured: Ethereum scaling networks hold $33.4 billion in total, of which $26.8 billion sits on rollups that post their data to Ethereum.
• A two-chain market: Base ($11.6 billion) and Arbitrum One ($10 billion) hold roughly 80% of all rollup value between them, both on optimistic designs.
• The largest validity proven chains: Mantle leads at $1.2 billion after its switch from fraud proofs, followed by Lighter at $849 million and Starknet at $370 million.
• Activity tells a different story: Lighter processed around 875 user operations per second over the past day, roughly nine times Base's 94, showing where fast finality wins.
• The long tail: L2Beat tracks 23 fully fledged rollups, and only a handful of small chains such as Facet and the privacy network Aztec have reached Stage 2, the final decentralization grade.

Ethereum Itself Is Going ZK

The strongest signal in the whole debate is what Ethereum is doing to its own base layer. In the middle of last year, the Ethereum Foundation set a target of proving 99% of mainnet blocks in under 10 seconds, fast enough for validators to verify a zero-knowledge proof instead of re-executing every transaction.

By December the target was met. Proving latency came down from 16 minutes to 16 seconds, costs fell 45-fold, and the roadmap has since moved on to hardening security before ZK clients reach mainnet.

The supporting upgrades are landing on schedule. Fusaka went live last December and introduced PeerDAS, which lets nodes verify data availability by sampling rather than downloading everything, and its follow-up blob forks more than doubled the data capacity rollups rely on.

The next hard fork, Glamsterdam, restructures block production so provers get several seconds to generate a proof instead of one or two.

Once Ethereum verifies validity proofs natively, fraud proof systems start to look like transitional technology, and every major optimistic stack has already put ZK proving somewhere on its roadmap.

Rollups Beyond Ethereum

The rollup model has also crossed over to Bitcoin, where it long seemed impossible because Bitcoin's scripting cannot verify complex proofs directly. BitVM changed that by allowing fraud disputes to be settled on Bitcoin without any protocol change, and it now underpins the first generation of trust minimized Bitcoin bridges.

Citrea put the idea into production in late January as the first live ZK rollup on Bitcoin. It executes EVM transactions, proves them with a RISC Zero based zkEVM, and inscribes the compressed proof to the Bitcoin blockchain, which serves as both its settlement and data layer.

Its Clementine bridge needs only one honest participant to keep withdrawals safe, a far stronger model than the federated custodians earlier Bitcoin sidechains depended on. The ecosystem is still tiny next to Ethereum's layer 2 networks, but the design questions being answered there are the same ones Ethereum worked through years earlier.

Bottom Line

ZK and optimistic rollups solved the same congestion problem from opposite ends, and after years of competition the market has delivered a split verdict. Optimistic chains won the adoption race, holding the overwhelming majority of value and applications, while zero-knowledge proofs won the argument about where the technology should end up.

For users, the choice today rests less on proof architecture than on what each chain offers around it. Fees are near identical since blobs arrived, so ecosystem depth, withdrawal timing, and a chain's L2Beat stage are the criteria we would weigh first. We expect the labels themselves to fade over the next few years as fraud proof chains keep bolting on validity proofs, and as Ethereum finishes turning its own base layer into the largest ZK verified system in crypto.