Ethereum’s rollup-centric roadmap has produced the deepest layer-2 ecosystem in crypto. Two years after the Dencun upgrade slashed blob costs by 10x, L2s now process more transactions per day than Ethereum mainnet itself. Here are the ten largest layer-2 networks by total value locked (TVL), updated for 2026. We focus on optimistic and zk-rollups settling to Ethereum and report Stage classification (per L2Beat) alongside ecosystem maturity, sequencer status, and the strength of native applications.
1. Arbitrum (ARB)
Arbitrum One remains the largest L2 by TVL, with Nitro’s fraud-proof rollup design and a deep DeFi ecosystem anchored by GMX, Aave V3, Camelot, and Pendle. The Arbitrum DAO controls one of the largest treasuries in crypto.
- Why it matters: Highest L2 TVL means deepest liquidity and the broadest DeFi composability outside of mainnet.
- Key risk: Sequencer centralization remains a long-running concern; Stage 1 status puts Arbitrum behind some peers on decentralization.
- Coverage: live profile · prediction
2. Optimism (OP)
Optimism pioneered the OP Stack — the shared development kit now powering Base, World Chain, Mode, and other Superchain members. Optimism Collective’s retroactive public goods funding (RetroPGF) is the most successful permissionless funding mechanism in crypto.
- Why it matters: OP Stack network effects: tooling, sequencer revenue sharing, and shared sequencing position Optimism as the most credible Ethereum-aligned L2 supercluster.
- Key risk: Centralized sequencer; fault proofs only shipped in 2024; bridge dependency on a 7-day challenge window for non-fast withdrawals.
- Coverage: live profile · prediction
3. Base (BASE)
Coinbase’s OP Stack L2 has become the on-ramp of choice for retail and the home of the on-chain social experiment (Friend Tech, Farcaster apps, on-chain games). Base regularly hits the highest daily transaction count of any Ethereum L2.
- Why it matters: Coinbase distribution + cheap blobspace = the consumer-crypto chain. Most new wallets are funded through Coinbase, then bridged to Base.
- Key risk: Single sequencer operated by Coinbase; regulatory exposure for Coinbase translates directly to Base. No native token yet — speculation about a future BASE token has been repeatedly denied.
4. zkSync Era (ZK)
Matter Labs’ zk-rollup uses a custom LLVM-based zkEVM that diverges from Type-1 equivalence in exchange for faster proving and a native account-abstraction-first design. The June 2024 ZK airdrop was one of the largest in L2 history.
- Why it matters: Account abstraction native, social recovery wallets, and gas paid in any token — zkSync’s UX bets are paying off for retail.
- Key risk: Type-4 EVM equivalence creates ongoing porting friction; the prover remains centralized; airdrop controversy left a portion of the community frustrated with sybil filters.
- Coverage: live profile · prediction
5. Starknet (STRK)
StarkWare’s validity rollup uses Cairo (a Turing-complete language for STARK proofs) and a non-EVM execution layer. Starknet leads the broader STARK ecosystem and was the first prover to land on mainnet in 2021.
- Why it matters: Strongest cryptography in the L2 set; STARKs are quantum-resistant; account abstraction is native and battle-tested via Argent.
- Key risk: Cairo isn’t Solidity — every dApp has to be rewritten. STRK token unlocks have pressured price; Stage 0 decentralization status.
- Coverage: live profile · prediction
6. Linea
ConsenSys’ Type-2 zkEVM is tightly integrated with MetaMask, Infura, and Truffle. Linea is the default chain MetaMask suggests to new users for cheaper Ethereum.
- Why it matters: Direct funnel from the largest Ethereum wallet (MetaMask has ~30M monthly active users) makes Linea a distribution play.
- Key risk: No native token has launched yet (rumored 2026); centralized sequencer; ConsenSys is a single point of organizational risk.
7. Scroll
Scroll is the open-source, EVM-bytecode-equivalent zkEVM. It uses a custom GPU prover and emphasizes minimal divergence from Ethereum semantics.
- Why it matters: Closest to pure Ethereum equivalence in the zk class; existing Solidity contracts deploy with no code changes.
- Key risk: Slower TVL growth than Linea and Base; team has historically been quiet on token timing; competing for the same EVM-equivalent niche as Polygon zkEVM and Taiko.
8. Polygon zkEVM (POL)
Polygon’s zkEVM is a Type-3 zk-rollup that anchors the broader Polygon 2.0 vision of a unified ZK-secured network including PoS chain migration and the AggLayer.
- Why it matters: AggLayer roadmap unifies liquidity across all Polygon chains; large enterprise integrations (Starbucks, Reddit collectibles) continue to drive traffic.
- Key risk: Identity crisis between Polygon PoS, zkEVM, CDK chains, and Miden; user-facing UX still fragmented.
- Coverage: live profile · prediction
9. Metis (METIS)
Metis pivoted from a fork of Optimism to a hybrid optimistic-ZK rollup design with a decentralized sequencer being progressively rolled out. It targets DAO-native applications.
- Why it matters: Decentralized sequencer is a genuine differentiator; lower fees than Arbitrum/Optimism with sufficient liquidity for mid-cap DeFi.
- Key risk: Thinner ecosystem outside of Hummus DEX and Maia; lower DAU than peer L2s; METIS has a small circulating supply that creates price volatility.
- Coverage: live profile · prediction
10. Mantle (MNT)
BitDAO’s Mantle is an OP Stack variant with modular DA via EigenDA. Mantle’s standout product is mETH liquid staking — its LST has become a top-10 ETH staking product.
- Why it matters: EigenDA integration cuts blob costs ~80% vs vanilla rollups; large treasury (~$3B) funds aggressive ecosystem grants.
- Key risk: EigenLayer dependency adds a new trust assumption; OP Stack alignment isn’t officially Superchain (yet); MNT inflation schedule from treasury is a watch item.
- Coverage: live profile · prediction
What to watch across L2s in 2026
Several themes will shape the L2 landscape over the next twelve months. First, native rollups — the proposed Ethereum primitive that would settle L2 state transitions directly within the L1 execution environment — would dramatically simplify the trust model but require consensus-layer changes. Vitalik’s “Surge” roadmap milestones include shared sequencing, generic ZK proofs for optimistic rollups, and improved blob throughput via PeerDAS.
Second, restaking continues to reshape the security narrative. EigenLayer-secured AVSs are now used by several L2s for data availability (EigenDA), watcher networks, and bridge attestations. The trade-off — additional slashing surface area in exchange for cheaper DA — is increasingly being chosen by new L2s aiming to minimize fee burden on users.
Third, the “Stage” framework from L2Beat continues to be a meaningful filter. Stage 0 chains (most L2s today) retain training-wheel features like upgrade keys held by multisigs. Stage 1 requires permissionless fault proofs and a robust exit window. Stage 2 — none yet — removes admin controls entirely. Expect at least two top-five L2s to advance to Stage 1 in 2026.
Finally, sequencer decentralization. Today, almost every major L2 runs a centralized sequencer operated by the L2 development company. Shared sequencing networks (Espresso, Astria, Radius) and proposer-builder separation for L2s are active research areas. The chain that delivers a credibly decentralized sequencer with low latency will gain a meaningful security narrative.
Methodology
TVL figures are pulled from L2Beat and DefiLlama and aggregated across the rollup’s primary smart contract suite (bridges, lending markets, DEX liquidity, LSTs deployed natively on the L2). We exclude wrapped assets that are also counted on mainnet to avoid double-counting. Stage classification follows the L2Beat decentralization framework (Stage 0 / 1 / 2). Sequencer status describes whether posting and ordering are still operated by a single party or by a permissioned set. Rankings are refreshed quarterly; this snapshot is current as of May 2026.
How these L2s compare to mainnet Ethereum
Mainnet Ethereum settles roughly 1.1 million transactions per day. The top ten L2s combined now exceed 15 million daily transactions, meaning rollups have decisively shifted the center of gravity for user activity. Average L2 fees range from ~$0.005 (Mantle, Base during normal congestion) to ~$0.05 (Arbitrum during peak demand) — versus $0.50-$5 on mainnet for similar transfers.
For builders, the choice between L2s now hinges on three factors: (1) where the target user base already holds liquidity (Base for Coinbase users, Arbitrum for DeFi natives, Solana-equivalent UX on zkSync), (2) what kind of dApps already exist there (DeFi blue chips on Arbitrum, consumer apps on Base, gaming experiments on Mantle and Linea), and (3) what the DA cost profile looks like for the specific workload.
Frequently asked questions about L2s
What is the difference between an optimistic and a ZK rollup?
Optimistic rollups (Arbitrum, Optimism, Base) post transaction batches to Ethereum and assume they are valid unless challenged within a fraud-proof window (typically seven days). ZK rollups (zkSync, Starknet, Linea, Scroll, Polygon zkEVM) instead post a cryptographic validity proof with every batch — there is no challenge period needed because the proof itself attests to the correctness of execution. ZK rollups have faster final settlement back to L1; optimistic rollups have historically had simpler execution environments and faster prover times.
Are L2 funds as secure as mainnet?
The security model of a rollup depends on three things: (1) the integrity of the proof system or fault-proof window, (2) the trustworthiness of the sequencer ordering transactions, and (3) the upgrade keys controlling the rollup’s smart contracts. Today, no major rollup is fully decentralized along all three axes. Users should treat L2 funds as having moderately higher risk than mainnet ETH, with the practical implication that long-term cold storage of large balances is still better done on L1.
Will L2s ever consolidate?
Probably not in 2026. The Superchain (OP Stack), Polygon AggLayer, and Arbitrum Orbit all enable application-specific or use-case-specific L2 spinoffs rather than consolidation around one chain. The likely trajectory is “many specialized L2s sharing liquidity via shared sequencing or AggLayer-style interop” rather than “everyone migrates to one winner.”
What is the simplest way to start using L2s?
For most users, the easiest on-ramp is Coinbase → Base (which has a native one-click bridge). For DeFi-native users, the typical path is Coinbase or Kraken withdrawal to MetaMask → bridge to Arbitrum or Optimism via the official bridges. Always use the official bridges, not third-party variants that surface in search results — bridge fraud is a recurring scam pattern.
For more rankings and analysis, see all analysis →