Ethereum Layer 2 Landscape: A Comparative Analysis
A rigorous comparison of Ethereum's leading Layer 2 scaling solutions across performance, security, decentralization, and cost metrics. This report evaluates optimistic rollups, ZK-rollups, and hybrid approaches to determine which architectures are best positioned for institutional and consumer adoption.
The Ethereum Layer 2 ecosystem has grown from a handful of experimental rollups to a diverse landscape of over 50 active networks processing a combined 12,000 transactions per second as of Q1 2026. This report provides a systematic comparison of the nine largest L2s by total value locked — Arbitrum, Base, Optimism (OP Mainnet), zkSync Era, Starknet, Linea, Scroll, Blast, and Polygon zkEVM — evaluating them across five dimensions: throughput, finality guarantees, decentralization, developer experience, and effective user cost. Our analysis reveals that no single L2 dominates across all dimensions, and the competitive landscape is likely to remain fragmented.
Throughput and cost metrics show meaningful variance across implementations. Arbitrum One leads in raw transaction volume, processing an average of 42 transactions per second with median gas costs of $0.003. Base, benefiting from Coinbase's distribution and the OP Stack's maturity, handles 38 TPS at similar cost levels. ZK-rollups present a different profile: zkSync Era achieves 28 TPS but offers cryptographic finality within approximately 20 minutes versus the 7-day challenge window for optimistic rollups. Post-Pectra, all L2s have benefited from expanded blob space, with average costs dropping 60% compared to pre-upgrade levels. Notably, several L2s have begun experimenting with alternative data availability layers (EigenDA, Celestia, Avail) to further reduce costs, though this introduces new trust assumptions.
Security architecture remains the most consequential differentiator among L2s. As of this writing, only Arbitrum and Optimism have achieved "Stage 1" decentralization as defined by L2Beat's risk framework, meaning they have functional fraud proof systems with a security council that can override only under strict conditions. zkSync Era and Starknet have deployed validity proofs that are verified on-chain, providing stronger mathematical guarantees, but both still rely on centralized sequencers and upgradeable contracts. The remaining L2s operate at "Stage 0," meaning their security ultimately depends on a trusted operator. For institutional adoption, this security spectrum is a critical consideration — a protocol handling billions in TVL with an upgradeable multisig represents a fundamentally different risk profile than one secured by mathematical proofs.
The developer experience across L2s has converged significantly as the ecosystem matures. All optimistic rollups based on the OP Stack or Arbitrum Nitro are fully EVM-equivalent, meaning existing Solidity code deploys without modification. zkSync Era supports Solidity through its custom compiler but introduces subtle differences in gas metering and precompile behavior that require testing. Starknet's Cairo language offers superior proof efficiency but requires developers to learn a new programming paradigm, creating a higher barrier to entry. Tooling availability — including block explorers, indexers, bridges, and oracles — is now comprehensive across the top five L2s but remains sparse for newer entrants.
The economic model of L2s has evolved beyond simple transaction fee collection. Arbitrum's ARB token governs a DAO treasury exceeding $4 billion, while the Optimism Collective's retroactive public goods funding has distributed over $200 million to ecosystem contributors. Base has adopted a unique model where a portion of sequencer revenue flows back to Optimism's collective treasury as part of the Superchain revenue-sharing agreement. These governance and economic structures will likely determine long-term ecosystem stickiness as much as raw technical performance.
Looking forward, several trends will reshape the L2 landscape. Based rollups — which use Ethereum L1 validators as sequencers — are gaining traction as a path to maximal decentralization and credible neutrality. Shared sequencing networks like Espresso and Astria promise to enable atomic cross-rollup transactions, potentially resolving the liquidity fragmentation problem. The emergence of L3s (application-specific chains settling on L2s) is creating a recursive scaling architecture, though critics argue this adds unnecessary complexity. Perhaps most significantly, the line between L2s and alternative L1s is blurring as sovereign rollups and validiums expand the design space beyond Ethereum-centric security models.
