Ethereum L2 activity consolidated in 2026. L2BEAT tracks 73 active rollups securing over $48 billion, and two chains hold 77% of that value. Distribution decided the winners. Base rides Coinbase. Arbitrum has Robinhood's Orbit chain. Kraken runs Ink on OP Stack. Chains without a distribution anchor are losing capital as incentive programs expire, and The Block's 2026 Layer-2 Outlook flags most mid-tier rollups as unlikely to survive the next 12 months.

AI agents transacting via agentic payment protocols like x402 are a small share of activity today. The chains that support them natively will compound that advantage.

Arbitrum: Liquidity and Composability

Arbitrum secures $14.9B to $16.9B in TVS as of May 2026 per L2BEAT, roughly 40% to 44% of the L2 market. It is the only rollup at L2BEAT Stage 1, meaning fraud proofs are permissionless.

DeFi liquidity runs deeper on Arbitrum than on any other L2. Aave, Uniswap, Curve, GMX, and Pendle all hold mature deployments. Native USDC arrived via Circle's CCTP in 2023, and combined stablecoin float on Arbitrum exceeded $4.2B in April 2026. Robinhood's Arbitrum Orbit chain, launched in late 2025, is the most visible institutional bet on the stack.

AI agents route through Arbitrum for the same reason DeFi teams do: depth. The x402 protocol supports Arbitrum via Permit2 for gasless approvals. Slippage rarely bites even on larger agent transactions. Projects building around DeFAI default here.

Best for: teams that need liquidity depth, mature tooling, and the security posture required for institutional partners or agent-driven traffic.

Base: Consumer Distribution and Growth

Base grew faster than any L2 in the past 18 months. TVL went from $2.1B in October 2024 to $11.2B by April 2026 per DefiLlama. On L2BEAT's TVS methodology, Base briefly overtook Arbitrum in Q2 2026. Activity metrics compound the case: 12.89 million daily transactions and 382,500 daily active users as of February 2026.

Coinbase is the moat. Its wallets, fiat rails, and account abstraction pipe mainstream users into Base with minimal friction. Aerodrome (the dominant ve(3,3) DEX), Morpho, and a long tail of consumer apps run there.

Base is also where agentic payments live. The x402 protocol was designed on Base, deployed there first, and still settles the majority of live volume there. See the B2B stablecoin payments guide for how this fits into broader payment infrastructure.

Best for: consumer products with a growth mandate, apps with fiat on-ramp requirements, agent-native products building on x402.

Optimism: The Superchain

Optimism as OP Mainnet holds $1.27B to $1.91B in TVS, well behind Arbitrum and Base. The direct number understates the position. OP Mainnet anchors the Superchain, which includes Base, Kraken's Ink L2, and Celo. The OP Stack collectively secures tens of billions.

Superchain interoperability matters for some architectures and is noise for others. Shared sequencing, native bridges across OP Stack chains, and coordinated upgrades give infrastructure teams a real edge when the product needs them. Retroactive Public Goods Funding and grants remain active.

Best for: projects targeting the OP Stack, teams pursuing grants, infrastructure protocols positioned to benefit from Superchain interoperability.

zkSync Era: zkEVM Under Mid-Tier Pressure

zkSync Era holds $165M to $405M in TVS depending on the source. That places it in the mid-tier band that saw net capital outflows through Q1 2026 as incentive programs expired. Its zkEVM prioritizes validity proofs and faster finality, and native account abstraction with paymaster patterns handles UX in ways optimistic rollups still cannot.

The technology is differentiated. Distribution is the problem. Without a Coinbase-style funnel or a Robinhood-style institutional anchor, zkSync earns users organically in a consolidating market. Teams considering it should test whether the ZK advantages (faster withdrawals, smaller proofs, better privacy primitives) matter for their specific product against Arbitrum's liquidity depth.

Best for: privacy-focused dApps, settlement-critical apps where finality speed beats liquidity depth, teams committed to ZK-native tooling.

Starknet: STARK Proofs and On-Chain Compute

Starknet holds approximately $617M in TVS. STARK proofs handle verification at scale, and Cairo is optimized for ZK computation. That delivers advantages for compute-intensive workloads (large NFT mints, complex on-chain proofs, batched-verification DeFi primitives) at the cost of a toolchain outside EVM workflows.

The ecosystem is smaller than the optimistic incumbents. The technical position is distinct. When a product's core value depends on cryptography or compute that STARKs handle better than any alternative, Starknet is correct. Otherwise the tooling cost is hard to justify against Arbitrum or Base.

Best for: projects requiring advanced cryptography, batched verification at scale, compute-intensive on-chain workloads.

Choosing an L2 in a Consolidated Market

Optimistic rollups (Arbitrum, Base, Optimism) hold roughly 80% of DeFi TVL and 77% of secured value in 2026. ZK rollups (zkSync Era, Starknet) hold technically differentiated but smaller positions.

Three criteria decide most deployments:

1. Where the users already are. Base and Arbitrum have the deepest activity. Deploying elsewhere means acquiring users from scratch.

2. Where the liquidity is. DeFi composability compounds on Arbitrum. Lending, swapping, and perps need counterparty depth that already exists.

3. How AI agents will use the product. Chains supporting x402 and stablecoin settlement natively capture that flow first. Today that means Base and Arbitrum.

The right L2 is the one that already holds the users and liquidity the product needs.