Ethereum's Layer 2 ecosystem holds approximately $48 billion in total value locked across 73 active rollups as of October 2026. Optimistic rollups — led by Arbitrum One, Base, and OP Mainnet — control roughly 80% of that capital, while ZK rollups including zkSync Era, Linea, Scroll, and Starknet ...
"In the longer term, ZK-Rollups are eventually going to beat Optimistic Rollups because they have these fundamental advantages like you don't need to have a seven-day withdrawal period. In more than 10 years from now or even more, I expect the Rollups to basically be all ZK." — Vitalik Buterin, Co-founder, Ethereum Foundation
Ethereum's Layer 2 ecosystem holds approximately $48 billion in total value locked across 73 active rollups as of October 2026. Optimistic rollups — led by Arbitrum One, Base, and OP Mainnet — control roughly 80% of that capital, while ZK rollups including zkSync Era, Linea, Scroll, and Starknet account for the remaining 20%. The gap has not narrowed in 2026; it has widened.
The data presents an uncomfortable paradox for ZK proponents. Optimistic rollups shipped earlier with near-native EVM compatibility, captured liquidity first, and built self-reinforcing network effects that ZK rollups have been unable to dislodge despite a theoretically superior security model. Base alone processes over 12 million daily transactions — roughly three times Arbitrum One's 4.3 million — and both dwarf the combined throughput of all major ZK rollups. Three networks (Base, Arbitrum, OP Mainnet) process approximately 90% of all L2 transactions and control roughly $33 billion of the ecosystem's TVL.
ZK rollups offer faster finality (minutes versus a seven-day challenge window), cryptographic security guarantees, and lower per-transaction data costs. None of these technical advantages has translated into market share. The question is no longer which technology is better. It is whether ZK rollups can overcome optimistic rollups' liquidity moat before the market fully consolidates around two or three dominant chains.
L2BEAT data as of late September 2026 shows Base at approximately $16.1 billion in TVL and Arbitrum One at $11.5 billion. Together these two chains secure $27.6 billion — 57% of all L2 value on their own. Adding OP Mainnet at $5.6 billion brings the optimistic rollup total to roughly $33.2 billion, or 69% of the $48 billion L2 ecosystem.
ZK rollup TVL is distributed across smaller positions: zkSync Era holds approximately $4.1 billion, Linea at $3.4 billion, Scroll at $2.1 billion, and Starknet at $1.5 billion. The combined ZK total of approximately $11.1 billion trails Base alone.
| Network | Type | TVL (Approx.) | Market Share | |---------|------|--------------|-------------| | Base | Optimistic | $16.1B | ~34% | | Arbitrum One | Optimistic | $11.5B | ~24% | | OP Mainnet | Optimistic | $5.6B | ~12% | | zkSync Era | ZK | $4.1B | ~9% | | Linea | ZK | $3.4B | ~7% | | Scroll | ZK | $2.1B | ~4% | | Starknet | ZK | $1.5B | ~3% | | Others (63 chains) | Mixed | ~$3.7B | ~7% |
The concentration extends to DeFi liquidity. Arbitrum One and Base together hold approximately 77% of all L2 DeFi liquidity. Combined stablecoin TVL across zkSync, Linea, and Scroll was $1.6 billion in April 2026 — roughly 20% of Arbitrum's stablecoin float alone, according to data tracked by DeFi aggregators.
Base leads on raw transaction count with 12.89 million daily transactions and 382,500 daily active users. Arbitrum One processes 4.3 million daily transactions with 129,000 daily active users. The disparity reflects Base's distribution advantage: direct integration with Coinbase's 100+ million user base funnels retail activity — social protocols, NFT markets, and small-value DeFi interactions — into high-frequency, low-value transactions.
ZK rollup transaction metrics are materially lower. Starknet's daily active addresses averaged approximately 20,000 as of mid-2026, down from over 100,000 during the pre-airdrop period in late 2023. The post-STRK airdrop collapse — active users dropping from roughly 300,000 to around 7,100 at the trough — demonstrated the extent to which activity had been incentive-driven rather than organic.
Linea has achieved broader reach through its MetaMask integration, claiming 30 million monthly active wallet users. However, wallet creation does not equal active on-chain engagement, and Linea's TVL of $3.4 billion suggests that most of those wallets hold minimal or zero balances.
The March 2024 Dencun upgrade introduced blob-carrying transactions via EIP-4844, reducing L2 data posting costs by 90-99%. This reshaped the economics for both rollup types.
Median transaction costs across major L2s now range from $0.02 to $0.06, down from approximately $0.25 pre-Dencun. Base's average fee per transaction stood at $0.00999 as of late September 2026; Arbitrum One's was $0.0221. A typical Uniswap-style swap costs $0.01 to $0.05 on Base and $0.02 to $0.10 on Arbitrum, depending on calldata size and L1 gas conditions.
ZK rollups theoretically hold a per-transaction data efficiency advantage once blob costs are factored in. Linea spent $0.012 per transaction in blob fees during Q1 2026, versus $0.018 for Arbitrum. However, this difference — fractions of a cent — is functionally irrelevant to users already paying near-zero fees on either architecture.
The blob base fee spent most of 2024 and 2025 at or near 1 wei, meaning rollups received near-free data availability while Ethereum L1 captured almost none of the value. This dynamic has shifted sequencer revenue to the primary source of L2 profitability.
Base generated approximately $93 million in sequencer revenue over 12 months, according to on-chain data analysis. Coinbase operates Base's sole sequencer, extracting sequencer margins near 85% during high-activity periods. Arbitrum operates with sequencer margins around 55%.
The profitability disparity underscores a structural tension: sequencer centralization — which critics describe as antithetical to blockchain's decentralization ethos — is the primary driver of L2 profitability. According to analysis published on BlockEden, Base's sequencer profit ratio stands at approximately 321:1 versus its L1 settlement costs, explaining why no L2 operator has voluntarily decentralized its sequencer.
This dynamic was exposed in June 2026 when Base's sequencer halted block production after processing an invalid block. The outage — the network's second sequencer-related incident — reignited criticism of single-operator dependency. L2BEAT assigns Base a Stage 1 decentralization rating, meaning fraud proofs are live but the sequencer remains under Coinbase's permissioned control.
ZK rollup sequencer economics are less transparent. zkSync Era, Linea, Scroll, and Starknet all operate centralized sequencers with no publicly disclosed revenue figures comparable to Base's on-chain data. None has decentralized its proving infrastructure.
The two rollup families employ fundamentally different security models:
Optimistic rollups assume transactions are valid and allow a seven-day challenge window for fraud proofs. Security relies on at least one honest verifier monitoring the chain and submitting challenges in time. Arbitrum One deployed permissionless fraud proofs in 2025; Base activated fraud proofs on the OP Stack the same year.
ZK rollups generate cryptographic validity proofs for each batch of transactions. The on-chain verifier contract checks the proof — if invalid, the batch is rejected outright. This model eliminates the assumption of an honest watcher: the math itself guarantees correctness, provided the proving system and verifier contract are sound.
The practical implication is withdrawal speed. Native withdrawals from optimistic rollups require waiting the full seven-day challenge period. ZK rollups complete withdrawals in minutes to hours once the validity proof is verified on L1. For institutional users with settlement-time requirements, or DeFi protocols requiring fast L1 composability, ZK rollups hold a structural advantage.
However, the market has largely mitigated the seven-day withdrawal delay through bridge protocols and liquidity networks that offer near-instant exits from optimistic rollups for a fee — typically 0.05-0.15% of the transfer value. This workaround has reduced the practical impact of the finality difference for retail users.
L2BEAT's staging framework classifies rollups from Stage 0 (centralized) to Stage 2 (fully decentralized):
| Network | Stage | Key Status | |---------|-------|-----------| | Arbitrum One | Stage 1 | Permissionless fraud proofs live | | Base | Stage 1 | Fraud proofs live, sequencer permissioned | | OP Mainnet | Stage 1 | Moving toward multi-proof redundancy | | Starknet | Stage 1 | First ZK rollup to reach Stage 1 (2025) | | zkSync Era | Stage 0 | Validity proofs active, sequencer centralized | | Linea | Stage 0 | Sequencer and prover centralized | | Scroll | Stage 1 | Validity proofs verified on L1 |
As of October 2026, no major L2 network has achieved Stage 2. Arbitrum is working toward Stage 2 with extended timelocks and multi-proof systems. Optimism is deploying Asterisc and Kona for multi-proof redundancy. Base follows OP Stack upgrades.
Among ZK rollups, Starknet was the first to reach Stage 1 in 2025. zkSync Era and Linea remain at Stage 0, a notable laggard status for chains that claim cryptographic security superiority. The irony is not lost on observers: ZK rollups' theoretical security advantage — validity proofs over fraud proofs — is undermined when the proving infrastructure itself remains centralized and upgradable by a single entity.
ZK rollups face three compounding disadvantages:
1. Liquidity cold start. DeFi follows liquidity, and liquidity follows users. Arbitrum captured DeFi's heavy hitters early — Aave, GMX, Uniswap — and built a self-reinforcing flywheel. ZK rollup DeFi ecosystems remain thinner. zkSync Era's top protocols (SyncSwap, Maverick, Aave v3) generate materially lower volume than their Arbitrum equivalents. Combined stablecoin liquidity across all major ZK rollups is roughly one-fifth of Arbitrum's alone.
2. EVM compatibility friction. Optimistic rollups achieved near-byte-level EVM compatibility from launch, allowing existing Ethereum contracts to deploy with minimal modification. ZK rollups required zkEVM implementations — complex engineering that introduced subtle incompatibilities, higher development costs, and longer deployment timelines. While zkSync Era, Scroll, and Linea have narrowed the compatibility gap, Starknet operates with Cairo, a non-EVM language that requires developers to learn an entirely new stack.
3. Proving costs. ZK proof generation is computationally intensive. While proving costs have declined with hardware optimization and protocol improvements, they remain a material overhead that optimistic rollups do not bear. This cost is either absorbed by the rollup operator (reducing margins) or passed to users (reducing competitiveness).
The counterargument — that ZK rollups will win "in the long run" because of superior finality and proof-based security — has been the consensus narrative since 2022. Four years later, the long run has not arrived. Ethereum co-founder Vitalik Buterin himself has stated that ZK rollups represent the long-term scaling endgame, but this view increasingly resembles a technical thesis disconnected from market dynamics.
The Ethereum L2 market as of October 2026 is not a technology competition. It is a liquidity and distribution competition, and optimistic rollups have won the current phase decisively.
ZK rollups retain a defensible technical thesis: faster finality, cryptographic settlement guarantees, and a more elegant long-term security model. For institutional settlement, compliance-sensitive applications, and cross-chain interoperability requiring proof verification, these properties may eventually command a premium.
However, "eventually" has been the operative word since 2022. Each quarter that optimistic rollups consolidate their TVL, developer ecosystem, and user base, the cost of switching for users and protocols rises. The L2 market is exhibiting the same winner-take-most dynamics visible in other platform markets: early distribution advantages compound, and technical superiority alone does not reverse them.
The 80/20 split between optimistic and ZK rollups may not be the final state. But the burden of proof has shifted. ZK rollups must now demonstrate not just that they are architecturally superior, but that this superiority matters to the marginal user or protocol choosing where to deploy $1 billion in liquidity. So far, the market's answer has been clear.