zkSync, Linea, and Scroll are the three largest production zkEVM rollups by TVL as of April 2026. For the broader L2 landscape including optimistic rollups, see the Ethereum L2 comparison guide. All three generate cryptographic validity proofs for every batch of transactions and post those proofs to Ethereum L1 for verification — meaning withdrawals to mainnet finalize in under 24 hours rather than the 7-day fraud-proof window that optimistic rollups require. Combined TVL across the three reached $9.6B in April 2026 per L2BEAT.
​

Each chain takes a different position on the EVM-equivalence trade-off. zkSync Era diverges from EVM bytecode for prover performance. Linea uses gnark-based proofs and aims for Type 2 zkEVM equivalence. Scroll prioritizes bytecode-level equivalence at the cost of prover throughput. This guide compares the three on proof system, fees, ecosystem depth, prover decentralization, and developer experience.
​

A zero-knowledge rollup is an L2 that batches transactions, executes them off-chain, and posts a cryptographic proof to Ethereum L1 attesting that the new state is valid. The proof — typically a zk-SNARK or zk-STARK — is mathematically verifiable in seconds on L1, regardless of how many transactions it covers. A user reading L1 state knows the L2's claimed balances and contract states are correct, with only the trust assumption of the underlying cryptography.
​

Compared to optimistic rollups (Base and Arbitrum are the dominant examples), ZK rollups eliminate the 7-day challenge window. Once a proof is verified on L1 (typically within 1–24 hours of the batch being posted), withdrawals finalize. Builders pay for this with prover infrastructure: generating proofs requires significant compute, and prover throughput is a real constraint on TPS for some designs.
​

The "zkEVM" subset of ZK rollups runs Ethereum-style execution. There are five formal zkEVM types per Vitalik's classification: Type 1 (fully Ethereum-equivalent), Type 2 (EVM-equivalent), Type 3 (mostly EVM-equivalent with limitations), Type 4 (Solidity-equivalent), Type 5 (custom VM). zkSync Era is Type 4, Linea is Type 2, Scroll is Type 2 with bytecode equivalence ambitions.
​

The proof system determines prover speed, finality time, and L1 verification cost. The three chains differ at every layer.

zkSync's Boojum upgrade in 2024 cut proof costs roughly 10x by switching from a custom system to redshift-based PLONK. The chain's zkEVM is Type 4, meaning Solidity contracts compile to a different bytecode (zkEVM IR) than Ethereum's EVM bytecode. This lets zkSync optimize the prover but requires recompilation for some contracts.
​

Linea uses ConsenSys's gnark-based proof system. Linea is Type 2 — EVM-equivalent at the execution layer, so most contracts deploy without modification. Proof generation runs on a centralized prover cluster operated by ConsenSys, with a roadmap to permissionless proving by year-end 2026.
​

Scroll uses Halo2 with KZG commitments. Its commitment to bytecode-level EVM equivalence means contracts deploy unmodified, and trace generation matches L1 EVM behavior closely. The trade-off is slower prover throughput — Scroll averages 14 TPS versus zkSync's 28.
​

ZK rollups have a fee structure with three components: L2 execution gas, L1 proof verification (amortized across all transactions in a batch), and L1 data availability via blobs. Per growthepie for the week ending April 24, 2026:

The fee gap between Linea and zkSync narrowed in Q1 2026 after Linea's blob compression upgrade landed. Scroll's higher fees reflect the cost of bytecode-equivalent proving — every EVM opcode must be encoded in the prover circuit, which is more expensive than zkSync's compiled IR.
​

Each chain attracts different protocols based on its proof system and developer experience.
​

zkSync Era ecosystem. Native account abstraction is built into the protocol — every account can be a smart contract without ERC-4337 overhead. This drove early adoption from wallet teams and a wave of paymasters. Top protocols: SyncSwap (DEX), Maverick Protocol, Aave v3, Woofi. Stablecoin TVL of $720M concentrates in USDC and USDT. The ZK token's 2025 airdrop drove a temporary usage spike that has since cooled.
​

Linea ecosystem. ConsenSys's MetaMask integration drives heavy retail wallet activity — Linea sees more unique daily wallets than its TVL ranking would predict. Top protocols: Renzo (LRT), ether.fi, EigenPie, Lynex (DEX), ZeroLend. Liquid staking and restaking dominate composition; LRT/ETH-yield protocols hold roughly 60% of Linea TVL.
​

Scroll ecosystem. The chain's bytecode-equivalence has made it the deployment choice for protocols that prefer not to modify contracts. Top protocols: Aave v3, Compound, Ambient Finance, Pencils Protocol. Smaller TVL but high protocol count — over 90 deployed dapps as of April 2026.
​

All three chains run centralized provers as of April 2026. The economics are challenging: a single proof can cost $50–500 in compute, and the prover operator earns the difference between user gas paid and L1 verification cost. Decentralizing means coordinating multiple operators while keeping proof latency low.
​

zkSync's roadmap calls for permissionless proving via the Boojum 2.0 upgrade, with a target window of late 2026. Matter Labs has published a multi-prover model where multiple operators stake ZK tokens to participate.
​

Linea has stated permissionless proving is on the 2026 roadmap but has not published specifics. ConsenSys currently operates the only production prover.
​

Scroll's prover is operated by Scroll Foundation. The chain has experimented with a decentralized prover network on testnet through partnerships with Gevulot and similar projects, but mainnet activation has no committed date.
​

Until proving decentralizes, the prover operator is a liveness dependency for each chain. If the operator goes offline, no new proofs are generated and L1 finality stalls (though L2 transactions still process via the sequencer).
​

Account abstraction is one of the practical differentiators across the three chains. zkSync Era ships native AA at the protocol level — every account can be a contract account without requiring ERC-4337 infrastructure. This means paymasters, session keys, and signature aggregation are first-class without bundlers. Linea and Scroll support ERC-4337 through standard bundler infrastructure (Pimlico, Stackup, Biconomy), but lack native protocol-level AA.
​

For wallet teams, zkSync's native AA lowers integration overhead. For dapps that just want gasless transactions or social-recovery wallets, ERC-4337 on Linea or Scroll works fine and benefits from a larger pool of bundler services. The choice usually comes down to whether the team builds a wallet (favor zkSync) or an application (any of the three).
​

The Type 4 vs Type 2 distinction matters for developers in three places. First, contract compatibility: zkSync requires recompiling Solidity through its zksolc compiler, which sometimes produces subtly different behavior for inline assembly or unusual opcodes. Linea and Scroll accept standard EVM bytecode. Second, gas estimation: zkSync uses a different gas pricing model that complicates direct ports of contracts hardcoding gas amounts. Third, debugging: standard EVM tracers work on Linea and Scroll; zkSync requires its own tracing tools.
​

For most teams writing standard Solidity, the practical experience on all three chains is similar — Hardhat, Foundry, Tenderly, and major indexers all support each. The exceptions matter for teams with complex contract logic, which is why some auditing firms charge a premium for zkSync deployments.
​

Bridging into and between ZK rollups uses a different cost profile — see the bridging cost guide for the full comparison — than optimistic rollups. Canonical bridges from L1 cost the same range ($4–18 in mainnet gas) but withdrawals back to L1 finalize within hours rather than 7 days. Third-party bridges complete in under 60 seconds with fees of 0.05–0.20% notional.
​

For zkSync Era, the most common bridges are zkSync Bridge (canonical, ~30 min withdraw), Across Protocol, and Orbiter Finance. Linea uses ConsenSys's native bridge plus Across, Stargate, and LayerZero. Scroll uses its native bridge plus Hop, Across, and Connext. CCTP integrations bring native USDC to zkSync Era today, with Linea and Scroll integrations in active development per Circle's roadmap.
​

For applications that need to move stablecoins from optimistic to ZK rollups (or between ZK rollups), the optimal path almost always uses a third-party bridge or stablecoin-native rail rather than the L1-canonical-then-L1-canonical-back path, which would take 7+ days end-to-end.
​

Stablecoin float is where ZK rollups still trail optimistic rollups. The L2 stablecoin liquidity guide ranks all major chains. Combined stablecoin TVL across zkSync, Linea, and Scroll was $1.6B in April 2026 — roughly 20% of Arbitrum's stablecoin float alone. Native USDC via CCTP launched on zkSync Era in Q1 2026 and is announced for Linea; Scroll currently uses bridged USDC.
​

For payments and treasury teams, this is the single biggest practical reason ZK rollups are not yet primary deployment targets. Liquidity for large USDC swaps or stablecoin DEX routes is materially thinner than on Arbitrum or Base. The gap is closing as native USDC integrations roll out, but as of April 2026 the optimistic rollups remain the default for stablecoin-heavy applications.
​

Eco supports zkSync Era, Linea, and Scroll as part of its 15-chain stablecoin execution network. Eco Routes (CLI + API) routes stablecoin transfers between any of these ZK rollups and the rest of the supported chains, abstracting the asymmetric finality times — ZK rollup withdrawals to L1 finalize faster than optimistic rollup withdrawals, but cross-chain hops still need a bridge or stablecoin-native rail. Eco selects the lowest-cost path: CCTP where supported, third-party rails like Across or Hyperlane otherwise. Teams expanding from optimistic rollups to ZK rollups for fast finality typically integrate Routes once rather than wiring per-chain bridges.
​

Linea has the lowest median fees among the three at $0.04 per USDC transfer in April 2026. zkSync Era is at $0.05 and Scroll at $0.06. Differences come from prover system efficiency and blob compression — Linea's recent compression upgrade narrowed the gap with the cheapest optimistic rollups.
​

Both inherit security from Ethereum L1, but the trust assumptions differ. ZK rollups assume the underlying cryptography (SNARKs or STARKs) holds. Optimistic rollups assume at least one honest validator runs a fraud-proof watcher during the 7-day challenge window. Both also depend on operator multisigs at current decentralization stages.
​

Linea and Scroll accept standard Solidity contracts without modification. zkSync Era requires recompiling through zksolc, which usually produces equivalent behavior but occasionally diverges for inline assembly or specific opcodes. Most contracts deploy unchanged across all three after the recompile.
​

Canonical withdrawals to Ethereum L1 finalize once a validity proof is verified — typically 30 minutes (zkSync), 75 minutes (Linea), or 90 minutes (Scroll). Third-party bridges like Across complete in under 5 minutes by fronting liquidity. CCTP withdrawals on chains with native USDC (zkSync, soon Linea) take 15–25 minutes.
​

zkSync's Type 4 zkEVM compiles to a custom IR optimized for proof generation, allowing higher prover throughput than Scroll's bytecode-equivalent design. Scroll's circuit must encode every EVM opcode exactly, which is more expensive per transaction. This is a deliberate trade-off — Scroll values bytecode equivalence; zkSync values throughput.
​

The two coexist with different strengths. ZK rollups offer faster L1 finality and don't require a fraud-proof challenge window. Optimistic rollups have larger ecosystems, deeper stablecoin liquidity, and more mature tooling. As ZK proving costs drop and ecosystems mature, the gap narrows, but optimistic rollups remain the default for most production DeFi today.
​

Linea is the most common entry point for teams already on Ethereum mainnet — Type 2 EVM equivalence and MetaMask distribution lower friction. Pick zkSync Era if your application benefits from native account abstraction or you want the highest TVL among ZK rollups. Pick Scroll if bytecode-level equivalence matters and you can absorb slightly higher fees.
​

Polygon zkEVM, Taiko, and Starknet are the next tier. Polygon zkEVM holds $250M TVL and is part of Polygon's AggLayer. Taiko is a Type 1 (fully Ethereum-equivalent) zkEVM with $180M TVL. Starknet uses STARK proofs and a custom Cairo VM rather than EVM, holding $1.5B TVL — covered in this guide's pillar comparison.
​