Ethereum Layer 2 Networks: The 2026 Definitive Guide

TL;DR

• Ethereum layer 2 networks (L2s) are separate blockchains that batch thousands of transactions, post compressed data back to Ethereum L1 as blobs (since the EIP-4844 Dencun upgrade of March 13, 2024), and inherit Ethereum's security via fraud proofs (optimistic rollups) or validity proofs (ZK rollups).
• As of mid-July 2026 the L2 ecosystem secures roughly $35B in TVS across 50+ live chains — down from ~$48B in May 2026 after crypto prices fell and L2BEAT removed ~$7B of manipulated, non-circulating RAIN token supply from Arbitrum's count on July 13. Base (~$11B) and Arbitrum One (~$10.4B) are now roughly level at the top; OP Mainnet (~$1.9B) is a distant third. Average L2 fees have fallen more than 90% since Dencun.
• The four winning frameworks are OP Stack (Optimism Superchain — 20+ chains incl. Base, Mode, Zora, World Chain), Arbitrum Orbit (100+ L3s), Polygon CDK + AggLayer, and ZK Stack (ZKsync Elastic Network).
• The frontier moves to based rollups, decentralized/shared sequencers (Espresso, Astria), alt-DA layers (Celestia, EigenDA, Avail), and the Stage 1 → Stage 2 decentralization push tracked by L2BEAT.

Table of contents

• What is an Ethereum Layer 2?
• Why L2s exist: Ethereum's scaling bottleneck
• How rollups actually work
• Optimistic rollups vs ZK rollups: side-by-side
• The L2 taxonomy: rollups, validiums, optimiums, sidechains
• Top L2s by TVL — mid-2026
• Base — Coinbase's consumer juggernaut
• Arbitrum — the DeFi heavyweight
• Optimism and the OP Superchain
• ZKsync Era and the Elastic Network
• Starknet, Scroll, Linea, Polygon zkEVM, Taiko, Blast, Mantle
• L2 frameworks: OP Stack, Orbit, CDK, ZK Stack, Caldera, Conduit, AltLayer
• Base vs Arbitrum vs Optimism vs ZKsync — the 2026 head-to-head
• L3s, AppChains, and the multi-chain future
• How to bridge to a Layer 2 — step-by-step
• L2 risks: sequencers, upgrade keys, bridge hacks
• The future: based rollups, shared sequencers, alt-DA, MEV redistribution
• FAQ
• Glossary

What is an Ethereum Layer 2?

An Ethereum layer 2 is a separate blockchain that runs on top of Ethereum and uses Ethereum as its settlement and data-availability layer. The term "Layer 2" specifically refers to networks that inherit Ethereum's security — anyone can reconstruct the L2 state from L1 data and force-exit funds even if the L2 operator vanishes. A sidechain like Polygon PoS trusts a separate validator set and is not a strict L2.

Settlement is the moment a transaction becomes irreversible from Ethereum's perspective; execution is the actual computation. L1 settles, L2s execute off-chain and send back a compressed proof or data batch. That is how rollups run 1,000+ TPS at a fraction of L1 cost — they compress 100 swaps into one Ethereum blob.

Modern L2s are almost all rollups. A rollup posts every transaction's data to Ethereum (full reconstructability) plus either (a) a fraud-proof system challengeable for 7 days — an optimistic rollup like Arbitrum, Optimism, or Base — or (b) a validity proof (zk-SNARK / zk-STARK) — a ZK rollup like ZKsync Era, Starknet, Scroll, or Linea.

Why L2s exist: Ethereum's scaling bottleneck

Ethereum L1 settles ~12–15 transactions per second. During the 2021 bull market gas fees spiked to $50–$200 per swap. Vitalik Buterin's Endgame post (2021) made the bet explicit: scale through rollups, keeping L1 a maximally decentralized settlement layer. By 2024 this was Ethereum's official rollup-centric roadmap.

Three milestones cemented the L2 era:

• The Merge (Sept 15, 2022) moved Ethereum to PoS, cutting energy ~99.95% and freeing block-space economics for scaling.
• Optimism's Bedrock (June 6, 2023) rewrote the OP Stack into a modular MIT-licensed framework — now used by Base, Mode, Zora, World Chain, and 20+ others.
• EIP-4844 / Dencun (March 13, 2024) introduced blobs. Within seven days L2 fees fell more than 90%. Base transactions dropped from ~$0.20 to under $0.02; Arbitrum swaps from $0.40 to $0.03.

Impact of EIP-4844 on Ethereum (May 2024) confirmed the fee collapse and quantified the new blob market. The follow-up 180 Days After EIP-4844 (Oct 2024) found blob demand scales sub-linearly — small L2s pay rent for unused blob space, motivating the shared-blob designs now appearing on the AggLayer and Espresso.

How rollups actually work

A rollup is a state machine that runs off-chain but commits to L1. The lifecycle:

1. User signs a transaction in their wallet (MetaMask, Coinbase Wallet, Rabby).
2. Sequencer receives it — today a centralized server run by Offchain Labs, Coinbase, Matter Labs, StarkWare, or ConsenSys — and gives a "soft" confirmation in 0.2–2 s.
3. Batch is built every few minutes, compressing thousands of transactions.
4. Batch is posted to L1 as a blob (EIP-4844) on the beacon chain. Blobs live ~18 days, long enough for watchers to reconstruct L2 state.
5. Proof is generated.
   • Optimistic rollup: claims the new state root is correct; a 7-day fraud-proof window opens. Optimism's Cannon VM (live since June 10 2024) is the reference.
   • ZK rollup: a prover generates a validity proof in minutes-to-hours; once verified on L1 the state is final.
6. Withdrawal. Optimistic rollups: 7-day wait or pay a fast bridge (Across, Hop). ZK rollups: minutes once proof finalizes.

Layer 2 Blockchains Simplified (2026) is the cleanest formal treatment; Towards a Formal Foundation for Blockchain Rollups (2024) is the canonical academic spec.

Optimistic rollups vs ZK rollups: the head-to-head

The 2024 paper Hollow Victory: How Malicious Proposers Exploit Validator Incentives in Optimistic Rollup Dispute Games is required reading on optimistic-rollup challenge-game economics; the 2024 paper Analyzing and Benchmarking ZK-Rollups and the 2025 paper Analyzing Performance Bottlenecks in Zero-Knowledge Proof Based Rollups are the references on ZK rollup performance.

Builder takeaway: ship on an optimistic rollup for maximum tooling and depth; ship on a ZK rollup for fast withdrawals, ZK-native gaming, or institutional flows that hate the 7-day exit.

The L2 taxonomy: rollups, validiums, optimiums, sidechains

L2BEAT classifies every "L2-like" chain by where the data lives:

• Rollup — data on Ethereum L1 as blobs. Highest security. Arbitrum, Base, OP, ZKsync, Linea, Scroll, Starknet, Taiko.
• Validium — data on off-chain DA committee. Validity-proof execution, weaker data security. Immutable X, Sorare, classic StarkEx.
• Optimium — optimistic execution + off-chain DA. Lowest fees, weakest security. Mantle (EigenDA), pre-2025 Manta Pacific.
• Plasma — deprecated predecessor design.
• State channel — niche (Connext, Raiden), not general-purpose.
• Sidechain — independent consensus, not strictly an L2. Polygon PoS, Gnosis Chain, Ronin.

A second axis is EVM equivalence (Vitalik's "different types of ZK-EVMs"): Type 1 = fully Ethereum-equivalent (Taiko, Scroll); Type 2 = minor optimizations (Polygon zkEVM); Type 2.5/3 = partial compatibility; Type 4 = high-level language only (Cairo/Starknet).

Top L2s by TVL — mid-2026

These are the largest rollup L2s by total value secured (TVS) per L2BEAT, cross-referenced against DefiLlama and The Block. Numbers are approximate and move daily — check L2BEAT's live rankings. Note: on July 13, 2026 L2BEAT removed roughly $7B of RAIN tokens from Arbitrum's TVS after concluding the supply was non-circulating and held in team-controlled multisigs ("heavily manipulated" per L2BEAT research); Arbitrum's figure below reflects the post-adjustment count.

Top-3 figures per L2BEAT/The Block (June–July 2026, Arbitrum after the July 13 RAIN adjustment); mid-table figures are May 2026 approximations (Starknet ~$617M, Linea ~$421M, ZKsync ~$404M per The Block). Polygon zkEVM, formerly a top-10 entry, was shut down on July 1, 2026 (see below). Newer chains — Unichain, Ink, Katana — now also rank in L2BEAT's top 12.

Aggregate L2 TVS peaked near $48B in May 2026 — up from $13B in March 2024 (pre-Dencun) — before falling to roughly $35B by mid-July 2026 on lower crypto prices and the RAIN methodology adjustment. Base and Arbitrum together account for roughly three-quarters of rollup TVS.

Base — Coinbase's consumer juggernaut

Base is the L2 operated by Coinbase, built on the OP Stack and launched August 9, 2023. Jesse Pollak (head of Base) and Brian Armstrong frame it as Coinbase's onchain platform — direct distribution to ~100M verified users.

The 2026 numbers are arresting:

• TVS: ~$11B (L2BEAT, mid-2026), up from $400M in early 2024 — now #1, roughly level with Arbitrum after L2BEAT's July 2026 RAIN adjustment.
• Revenue: the highest of any L2 in 2025, and the only major L2 to turn a profit — roughly $55M after L1 data costs and the Optimism Collective revenue share.
• DeFi share: ~46% of all L2 DeFi TVL flows through Base. Aerodrome is the dominant DEX, often #1 by daily fees across Ethereum and L2s.
• Daily transactions: 8–12M, regularly leading all L2s.
• Stage 1: reached April 29, 2025 (fault proofs + security council) — the 10th chain to reach at least Stage 1 per L2BEAT.
• cbBTC: Coinbase's wrapped Bitcoin natively on Base — $3B+ supply by April 2026 — is the default BTC representation in Ethereum DeFi.

Base has no native token. Jesse Pollak's September 2025 reframing opened the door to a possible future network token, but Coinbase has not committed to a launch.

Arbitrum — the DeFi heavyweight

Arbitrum One is the flagship rollup of Offchain Labs, launched August 31, 2021 and governed by the Arbitrum DAO since the ARB airdrop of March 23, 2023.

By April 2026:

• TVS: ~$17B through June 2026. On July 13, 2026 L2BEAT removed roughly $7B of non-circulating, team-controlled RAIN tokens from the count — calling the token "heavily manipulated" — cutting Arbitrum's TVS to ~$10.4B, roughly level with Base.
• DeFi share: ~31% of L2 DeFi liquidity. GMX, Pendle, Camelot, Aave V3 together do $1B+ daily volume.
• Stylus: WASM environment for Rust, C, and C++ smart contracts alongside Solidity. May 2024 mainnet; Q4 2025 Stylus Sprint allocated 5M ARB in grants.
• Timeboost (April 2025): optional MEV auction for transaction ordering. By Q1 2026 it had routed $2M+ to the Arbitrum DAO — the first L2 token directly capturing fee flow.
• Arbitrum Orbit: 100+ chains incl. Xai (gaming), ApeChain, Sanko, Proof of Play. Orbit chains settle on Arbitrum One.
• Arbitrum Nova: AnyTrust DA-committee chain for high-throughput gaming and social.
• LTIPP: 50M ARB Long-Term Incentive Pilot Program has bootstrapped DEX, perp, and RWA activity since 2024.

Optimism and the OP Superchain

OP Mainnet is the original optimistic rollup, operated by OP Labs. The OP Superchain federates 25+ chains running the OP Stack: Base, Mode, Zora, World Chain, Soneium (Sony), Unichain (Uniswap), Ink (Kraken), Bob, and many more.

Three differentiators:

1. Stage 1 leadership — Reached June 10, 2024 with permissionless Cannon fault proofs. Any third party can challenge an invalid state root.
2. OP Stack as public good — MIT-licensed and the most-forked L2 codebase in history.
3. RetroPGF — Retroactive Public Goods Funding distributes a fraction of sequencer revenue. Round 5 (October 2024) allocated 8M OP to OP Stack contributors; Round 6 (Q4 2024) allocated 2.4M OP to governance contributions; Retro Funding continued through 2025 under Season 7's metrics-based missions.

Interop (ERC-7802) went live across Superchain mainnets in early 2026, enabling one-block bridging between Base, OP Mainnet, Mode, and other Superchain peers.

ZKsync Era and the Elastic Network

ZKsync Era from Matter Labs launched March 2023. It uses Boojum (custom STARK prover) and a Type-2.5 zkEVM.

Pivotal events:

• ZK token airdrop (June 17 2024) — 17.5% of 21B ZK supply to 695,232 wallets; controversial sybil density and a $0.30 → $0.10 decline within months.
• April 15 2025 admin-key hack — a compromised admin called sweepUnclaimed() and minted 111M ZK (~$5M). Matter Labs negotiated a 90% return for a 10% bounty; ~$5.7M returned by April 23. The most-cited example of why upgrade keys are an L2's biggest single risk.
• Elastic Network / ZK Stack — Matter Labs's framework for ZK-secured L2s/L3s with shared liquidity. Adopters: Cronos zkEVM, GRVT, Sophon, Lens.

ZKsync pioneered native account abstraction (every account is a smart contract) and paymasters (apps pay user gas in any token), making it a favorite for consumer wallet teams.

Starknet, Scroll, Linea, Polygon zkEVM, Taiko, Blast, Mantle

Starknet

Starknet by StarkWare is the largest non-EVM ZK rollup, using Cairo and zk-STARKs. In 2025 Starknet became the first ZK rollup to reach Stage 1 on L2BEAT's decentralization ladder, and in late October 2025 it became the first L2 to integrate Bitcoin staking — a dual-token consensus in which STRK carries 75% and staked BTC 25% of the network's security weight, with BTC stakers (via wrappers like WBTC and tBTC) earning STRK rewards, anchoring L2 security to both Ethereum and Bitcoin.

Scroll

Scroll, operated by the Scroll Foundation, is a Type-2 zkEVM bytecode-compatible with Ethereum. The SCR token launched October 22, 2024 with a 7% airdrop (70M tokens). By mid-2026 Scroll had slipped out of L2BEAT's top-10 TVS rankings amid the broader ZK-cohort decline; flagship DEXs are SyncSwap, Ambient, and Zebra.

Linea

Linea, built by ConsenSys and tightly integrated with MetaMask, drove $1B+ TVS via "Voyage" and "Surge" campaigns. Q1 2026 distributed ~$30M in LINEA rewards across DeFi protocols. MetaMask Card gives Linea the most direct fiat onramp of any L2.

Polygon zkEVM and AggLayer

Polygon zkEVM is a Type-2 zkEVM, but Polygon's strategic bet is the AggLayer — a unified bridge and message-passing protocol turning any chain (Ethereum L2, CDK, Solana plug-ins) into a node in one liquidity graph. The MATIC → POL migration (September 4, 2024) made POL the staking token for Polygon 2.0. Polygon announced the zkEVM sunset in June 2025 and shut the network down on July 1, 2026, consolidating around Polygon PoS and the AggLayer (a claims interface for stranded EOA funds runs through December 2027).

Taiko

Taiko Alethia launched mainnet on May 27, 2024 — the first based rollup in production. Sequencing is delegated to L1 proposers, eliminating sequencer-centralization risk. The TAIKO token launched June 2024 with a 5% genesis airdrop.

Blast

Blast launched November 2023 with native ETH and stablecoin yield (4–5% APY) routed via Lido and MakerDAO. Peak TVS hit $2.7B; the June 2024 BLAST airdrop attracted controversy for vesting locks and referral mechanics. Q1 2026 TVS had retraced to ~$350M with persistently negative protocol revenue.

Mantle

Mantle runs an OP Stack fork using EigenDA for data availability (technically an optimium). Killer app: liquid staking via mETH and Mantle Restaking ($1B+ restaked ETH). MNT captures sequencer revenue via DAO buybacks.

L2 frameworks: OP Stack, Orbit, CDK, ZK Stack, Caldera, Conduit, AltLayer, Movement

By 2026, "deploying an L2" means picking a framework. The leading stacks:

• OP Stack — Optimism's MIT-licensed framework. Powers Base, Mode, Zora, World Chain, Sony's Soneium, Uniswap's Unichain, Kraken's Ink, Lyra, Cyber, Redstone, Mint, Orderly, Lisk, Bob, Snaxchain, and 10+ more, federated as the OP Superchain.
• Arbitrum Orbit — 100+ chains, mostly L3s settling on Arbitrum One. Notable: Xai, ApeChain, Sanko, Proof of Play, Treasure.
• Polygon CDK + AggLayer — ZK-native stack with unified liquidity. Adopters: Astar zkEVM, X1 (OKX), Canto.
• ZK Stack / Elastic Network — Matter Labs's framework. Adopters: Cronos zkEVM, Sophon, GRVT, Lens.
• Starknet Stack (Madara appchains) — Cairo L3s, gaming-focused.
• Caldera — RaaS, 70+ rollups (Manta, Kinto, RARI), multi-stack.
• Conduit — RaaS for OP Stack + Orbit (Proof of Play, Aevo, Lyra, Mode).
• AltLayer — restaked rollups using EigenLayer for fast finality (Cyber, Polyhedra).
• Movement — Move-on-EVM L2 from Movement Labs.

The 2024 paper The Espresso Sequencing Network is the foundational reference for shared sequencing atop these frameworks.

Base vs Arbitrum vs Optimism vs ZKsync — the 2026 head-to-head

Decision tree: consumer/social/payments → Base; DeFi, perps, RWAs → Arbitrum; decentralization and public goods → Optimism; ZK primitives, account abstraction, Move/Cairo → ZKsync, Starknet, or Movement.

L3s, AppChains, and the multi-chain future

L3s settle on an L2 instead of L1, paying L2 blob fees that the L2 batches further before posting to Ethereum — ~100x compression on cost. The big L3 ecosystems in 2026:

• Arbitrum Orbit — 100+ chains: Xai, ApeChain, Sanko, Proof of Play, Treasure, Kinto, Degen Chain.
• OP Stack L3s — World Chain (Worldcoin, via Conduit), Arena-Z, Forma.
• ZK Stack L3s — Cronos zkEVM, GRVT, Sophon, Lens (Aave social).
• App-specific chains (adjacent, not strict L3): dYdX v4 (Cosmos), Hyperliquid L1 (perps, $1B+ daily volume), Frax fraxtal (OP Stack), Aevo (Conduit OP Stack).

Cross-Rollup Atomic Transaction Execution (CRATE) and Cross-Rollup MEV describe the cross-chain MEV and atomicity problems L3 proliferation creates — and the shared-sequencer / preconfirmation designs that solve them.

How to bridge to a Layer 2 — step-by-step

The cheapest way is rarely the fastest. Decision matrix:

1. CEX withdrawal — Coinbase, Binance, Kraken, OKX support direct withdrawal to most L2s. Free, instant; best for first-time funding.
2. Canonical bridge — Base Bridge, Arbitrum Bridge, Optimism Bridge, Linea Bridge, Starknet Bridge, Mantle Bridge. Maximum security; 7-day exit on optimistic rollups.
3. Intent-based bridges — Across leads in 2026 (UMA-secured solver network). ~0.05–0.10% + L1 gas, 1–2 min.
4. LayerZero / Stargate — Stargate is the largest; more long-tail routes, higher fees.
5. Hop — Hop is the original L2↔L2 AMM bridge; battle-tested.
6. Aggregators — Squid (Axelar) and LI.FI route across many bridges.
7. Circle CCTP V2 — CCTP V2 burns and mints native USDC. Zero slippage, ~30s. Live on Ethereum, Arbitrum, Base, OP, Polygon, Avalanche, Solana.

Security tip: verify the bridge contract against L2BEAT's official-bridges page. Cross-chain bridges have lost more user funds than any other DeFi vertical — Ronin ($625M, 2022), Wormhole ($325M, 2022), Nomad ($190M, 2022), Multichain ($231M, 2023), Orbit Bridge ($82M, 2024).

L2 risks: sequencers, upgrade keys, bridge hacks

The 2024–25 academic and incident record points to five primary L2 failure modes:

1. Sequencer centralization. Every major L2 except Taiko (based) and dYdX v4 (Cosmos validator set) runs a single sequencer that can censor, reorder, or briefly halt the chain. Force-inclusion mechanisms let users push transactions directly to L1 if censored. Decentralized-sequencer roadmaps exist for Arbitrum (BoLD, Espresso), Base (announced 2026), Optimism (Superchain shared sequencing), and ZKsync.
2. Upgrade keys / security council compromise. Every Stage 0/1 rollup has a security council that can upgrade contracts on a 0–7-day delay. The April 2025 ZKsync admin-key incident proved the risk is operational. Mitigations: hardware-key M-of-N multisigs, timelocks, eventual Stage 2 transition.
3. Bridge hacks. ~$2.5B+ stolen from cross-chain bridges since 2021 — the single largest crypto loss vector. Canonical bridges are usually safe; third-party fast bridges have an incident history.
4. Liquidity fragmentation and bad MEV economics. Liquidity Fragmentation or Optimization? shows the same asset on five L2s causes 30–60% wider spreads vs. consolidated L1 liquidity. Rolling in the Shadows and When Priority Fails: Revert-Based MEV on Fast-Finality Rollups document MEV invisible on L1 but extractive on L2.
5. Token unlocks and governance attacks. ARB, OP, STRK, ZK, SCR, TAIKO all face multi-year unlocks pushing 30–80% additional supply through 2027. DAO governance attacks (Compound's GoldenBoys raid attempt, July 2024) remain a live threat.

The future: based rollups, shared sequencers, alt-DA, MEV redistribution

Five vectors define the L2 frontier through 2027:

• Based rollups — Proposed by Justin Drake in 2023; Taiko is the first mainnet. Outsource sequencing to L1 proposers for native MEV redistribution. Preconfirmations soften L1's 12-second block time.
• Shared sequencers — Espresso (HotShot consensus + Tiramisu DA) began its phased mainnet rollout with Mainnet 0 in November 2024, scaling toward permissionless participation from 2025; HotShot finalizes in seconds. Astria and AltLayer MACH offer alternatives. Adoption: Caldera and other RaaS chains.
• Alt-DA layers — Celestia (mainnet Oct 31 2023), EigenDA (live April 2024 on $20B+ restaked ETH), Avail (late 2024). Cuts L2 cost ~10x but adds a separate trust assumption.
• MEV redistribution — Arbitrum's Timeboost (April 2025), Flashbots's Rollup-Boost on Unichain, and PBS-style proposer auctions are starting to capture MEV that previously leaked to private searchers. Expect this to be a major L2 revenue line by 2027.
• Stage 2 decentralization — No major L2 had reached Stage 2 by mid-2026 (only a few small chains qualify per L2BEAT), but Arbitrum's BoLD, OP Stack's permissionless fault proofs, and ZKsync's prover-network put Stage 2 in reach for the leaders by 2027.

Towards Cost-Effective ZK-Rollups is the canonical reference on the proving-cost curve, and A Practical Rollup Escape Hatch Design is the blueprint for forced-exit guarantees.

FAQ

What is an Ethereum Layer 2?

An Ethereum Layer 2 (L2) is a separate blockchain that executes transactions off Ethereum mainnet but inherits Ethereum's security by posting transaction data and proofs back to L1. Rollups batch thousands of transactions and post compressed data as blobs (since EIP-4844). Fraud proofs (optimistic rollups) or validity proofs (ZK rollups) ensure the L2 cannot lie about its state. Users get Ethereum-grade security with 10–100x lower fees and 10–50x higher throughput. Top L2s in 2026 include Base, Arbitrum One, OP Mainnet, Mantle, Starknet, Linea, and ZKsync Era.

Optimistic rollup vs ZK rollup — which is better in 2026?

Optimistic rollups (Arbitrum, Optimism, Base, Blast) rely on a 7-day fraud-proof challenge window. ZK rollups (ZKsync, Starknet, Scroll, Linea, Polygon zkEVM) post a validity proof every batch — withdrawals finalize in minutes. Optimistic rollups have higher TVL, more mature tooling, and full EVM equivalence; ZK rollups offer near-instant finality and stronger security. By 2026 the gap has narrowed as Type-1 zkEVMs run Solidity bytecode unchanged and proving costs have dropped ~10x in two years.

What is the best Ethereum L2 in 2026?

There is no single best L2. By total value secured, Base (~$11B) and Arbitrum One (~$10.4B after L2BEAT's July 2026 removal of ~$7B in manipulated RAIN tokens) are roughly level at the top, with OP Mainnet (~$1.9B) third. Base wins for consumer apps and Coinbase distribution; Arbitrum dominates DeFi and perps (GMX, Pendle); Optimism leads on decentralization (Stage 1 since June 2024); ZKsync Era and Starknet lead for ZK-native apps; Scroll and Linea among Type-1/Type-2 zkEVMs; Taiko architecturally as the first based rollup.

How did EIP-4844 (Dencun) reduce L2 fees?

EIP-4844 ("proto-danksharding") shipped March 13, 2024 in the Dencun upgrade. It introduced "blobs" — a cheap data layer rollups use to post compressed transaction data. Before Dencun rollups paid for permanent calldata; after, they pay for blob space verified once and discarded after ~18 days. Within a week, average L2 fees fell more than 90%. Base dropped from ~$0.20 to under $0.02 per transaction.

What is L2BEAT and why are "stages" important?

L2BEAT tracks every L2's total value secured (TVS) and decentralization maturity. Its Stages framework: Stage 0 = full training wheels (council can override state); Stage 1 = permissionless proofs live with council backstop; Stage 2 = fully decentralized. As of mid-2026 more than ten chains have reached at least Stage 1 (Arbitrum, OP Mainnet, Base, and Starknet among them); only a few small chains have reached full Stage 2.

How do I bridge to a Layer 2?

Use the canonical bridge (Base Bridge, Arbitrum Bridge, Optimism Bridge, Linea Bridge, Starknet Bridge) for maximum security — 7-day exits on optimistic rollups. For fast transfers, use Across (intent-based), Stargate (LayerZero), Hop, Squid, or LI.FI. For native USDC, Circle's CCTP V2 burns and mints across chains in under a minute with no slippage. Always verify bridge addresses against project docs and L2BEAT.

What is the difference between an L2 and a sidechain?

An L2 inherits Ethereum's security: state is reconstructable from L1 data and users can force-exit even if the sequencer dies. A sidechain (Polygon PoS, Gnosis Chain, Ronin) has its own validator set with independent trust assumptions. The Ronin hack ($625M, March 2022) is the canonical sidechain-risk example. Polygon retired its zkEVM rollup in July 2026 and now centers on the Polygon PoS sidechain and the AggLayer.

What is a based rollup?

A based rollup outsources sequencing to Ethereum L1 validators. The L1 proposer for each slot decides the L2 transaction order, so the L2 inherits Ethereum's liveness and gets MEV redistribution by default. Taiko launched the first based rollup mainnet on May 27, 2024. Espresso, Booster Rollups, and "based + preconfirmations" designs add sub-second UX on top.

Are L2 native tokens (ARB, OP, STRK, ZK) worth holding?

Most L2 tokens are governance-only. Arbitrum's Timeboost (April 2025) is the first to capture direct sequencer revenue — over $2M to the Arbitrum DAO in its first months. Optimism's RetroPGF distributes a fraction of fees to public-goods projects. Base, ZKsync, and Linea route 100% of sequencer revenue to the operator. Token unlocks remain a major overhang. Treat L2 tokens as governance bets, not yield-bearing claims.

What are the biggest risks of using Layer 2s?

Five risks: (1) sequencer centralization (single sequencer can censor or reorder); (2) upgrade keys (council compromise can drain bridge — ZKsync April 2025 admin-key incident minted 111M ZK tokens); (3) bridge hacks (historically the largest loss vector, $2B+ stolen since 2021); (4) liquidity fragmentation across many L2s; (5) token unlocks and governance attacks remain ongoing across the ecosystem.

Glossary

• Blob (EIP-4844) — Temporary data attached to Ethereum blocks (~125 KB each, ~18-day retention) that rollups use to post compressed transaction data. Up to 9 per block (target 6) in April 2026.
• Canonical bridge — The official L1↔L2 bridge run by the rollup. 7-day withdrawal on optimistic rollups, minutes on ZK.
• Cannon — Optimism's MIPS-based fault-proof VM, live on OP Mainnet since June 10 2024.
• DA (data availability) — Guarantee that rollup data is published somewhere reconstructable. Rollups: Ethereum L1. Validiums: Celestia, EigenDA, Avail.
• EVM equivalence — How closely an L2's execution matches Ethereum L1 (Type 1 full → Type 4 high-level only).
• Fault / fraud proof — Cryptographic challenge disputing an invalid state root on an optimistic rollup.
• OP Stack — Optimism's MIT-licensed L2 framework powering 25+ chains.
• Preconfirmation — Near-instant promise from an L1 proposer to include a transaction in a future block.
• Sequencer — Actor that orders L2 transactions and produces L2 blocks.
• Stage 0/1/2 — L2BEAT's decentralization rubric.
• Stylus — Arbitrum's WASM env for Rust, C, C++ contracts alongside Solidity.
• Superchain — Federation of OP Stack chains sharing security and (eventually) sequencing.
• Timeboost — Arbitrum's MEV-ordering auction (April 2025) routing fees to the DAO.
• Validity proof — zk-SNARK/STARK proving an L2 state transition is correct.
• Validium — ZK-secured L2 that posts data off-chain. Cheaper, weaker security.

Related reading

• What is EIP-4844? Blobs explained
• Zero-knowledge proofs explained
• Optimistic rollups explained
• Data availability: Celestia, EigenDA, Avail
• Ethereum staking 2026 guide
• DeFi TVL explained
• MEV on Ethereum and L2s

Sources and further reading

• L2BEAT — State of the layer two ecosystem (https://l2beat.com)
• L2BEAT — Total Value Secured rankings (https://l2beat.com/scaling/tvs)
• growthepie.xyz — L2 daily activity dashboards (https://growthepie.xyz/)
• DefiLlama — Rollup chains TVL (https://defillama.com/chains/Rollup)
• Ethereum Foundation — Dencun mainnet announcement (https://blog.ethereum.org/2024/02/27/dencun-mainnet-announcement)
• Ethereum.org — Layer 2 overview (https://ethereum.org/en/layer-2/)
• Optimism — Permissionless Fault Proofs and Stage 1 Arrive to the OP Stack (https://www.optimism.io/blog/permissionless-fault-proofs-and-stage-1-arrive-to-the-op-stack)
• OP Stack docs — Cannon fault proof VM (https://docs.optimism.io/op-stack/fault-proofs/cannon)
• Arbitrum docs — Stylus quickstart (https://docs.arbitrum.io/stylus/stylus-quickstart)
• ZKsync docs — ZK Stack (https://docs.zksync.io/zk-stack)
• Starknet — Decentralization Roadmap (https://www.starknet.io/roadmap/)
• Polygon — MATIC → POL migration (https://polygon.technology/blog/save-the-date-matic-pol-migration-coming-september-4th-everything-you-need-to-know)
• Scroll — SCR token launch (https://scroll.io/blog/scr-token)
• Taiko — Project home (https://taiko.xyz/)
• Espresso Systems (https://www.espressosys.com/)
• Astria — Shared Sequencer Network (https://www.astria.org/blog/astria-the-shared-sequencer-network)
• Celestia (https://celestia.org/)
• EigenDA (https://www.eigenda.xyz/)
• Base docs (https://docs.base.org/)
• Vitalik Buterin — Endgame (https://vitalik.eth.limo/general/2021/12/06/endgame.html)
• Vitalik Buterin — The different types of ZK-EVMs (https://vitalik.eth.limo/general/2022/08/04/zkevm.html)
• The Crypto Times — L2Beat calls RAIN "heavily manipulated" after $7B TVS cut (July 14, 2026) (cryptotimes.io)
• The Block — 2026 Layer 2 Outlook (theblock.co)
• Base — Base has reached Stage 1 decentralization (April 29, 2025) (blog.base.org)
• Cointelegraph — Starknet hits Stage 1 decentralization, tops ZK-rollups for value locked (cointelegraph.com)
• The Block — Starknet introduces Bitcoin staking (theblock.co)
• The Crypto Times — Polygon shuts down its once-flagship $250M zkEVM on July 1 (cryptotimes.io)
• Espresso Systems — Espresso Mainnet 0 is live (medium.com)
• Optimism — Retro Funding 6: Governance round details (gov.optimism.io)

About the author

DeFi Intel Research is an independent research desk covering rollup architectures, MEV economics, and onchain market microstructure. Our analysts maintain live infrastructure across Arbitrum, Base, OP Mainnet, ZKsync Era, Linea, Scroll, Starknet, and Taiko. Author profile: /about.

Last updated: 2026-07-15