The Future of Gas: Who Pays on a Dozen Different L2s?

Gas is no longer singular. Ethereum's L1 gas market is a single, global auction. Every L2—Arbitrum, Optimism, zkSync, Base—operates its own independent fee market with unique pricing logic, creating a fragmented user experience.

The payer is not always the user. Account abstraction and sponsored transactions shift gas payment to dApps or wallets, turning gas from a user cost into a customer acquisition tool. Protocols like Biconomy and Pimlico enable this.

Fee predictability is dead. On L1, gas is volatile but the model is simple. On L2s, fees are a composite of L1 data posting costs, local execution, and sequencer/prover margins, making cost estimation a protocol-specific calculation.

Evidence: A user bridging from Arbitrum to Polygon zkEVM via Across pays fees in three different assets across two distinct fee models, a process no normal user can mentally model.

Gas payment is UX fragmentation. Every new L2 introduces its own native token or ETH-based fee market, forcing users to manage a dozen different balances just to transact, a problem that scales with the number of chains.

The winning abstraction will subsidize gas. Protocols that solve this—like account abstraction via ERC-4337 or intent-based systems like UniswapX—will win by making gas invisible, shifting the cost to applications or solvers.

This creates a new business model. Relayers, bundlers, and solvers become critical infrastructure, monetizing gas sponsorship and order flow, similar to how Coinbase or MetaMask currently capture wallet-based revenue streams.

Evidence: Arbitrum processes over 1M daily transactions, yet users still need ARB or bridged ETH to pay for them; this friction directly limits cross-chain activity and composability.

Fee abstraction is the new standard. Users now expect to pay for L2 gas with any token, a UX paradigm pioneered by zkSync's native account abstraction and Arbitrum's gas token support. This shifts complexity from the user to the protocol's gas subsidy logic.

The sequencer pays, you don't. Chains like Base and Optimism use a 'transaction sponsorship' model where apps prepay ETH to the sequencer, creating a seamless, gasless experience for end-users. This turns gas from a user cost into a developer acquisition cost.

Proof systems dictate final cost. A ZK-rollup's gas includes a fixed cost for proof generation and verification, while an Optimistic rollup's cost is dominated by L1 data posting. This creates divergent long-term cost curves as proof hardware improves.

Evidence: Starknet's recent fee reduction of 50% was achieved not by lowering L1 costs, but by optimizing their Cairo-VM execution and STARK prover efficiency, demonstrating that L2 gas is now a software optimization problem.

Gas abstraction is a business strategy. Layer 2s like Arbitrum, Optimism, and Base subsidize user transactions to drive adoption and lock in volume. This creates a zero-friction onboarding experience where the protocol or a sponsor pays, not the end-user. The cost is a marketing expense offset by sequencer revenue and increased TVL.

The payment model dictates the abstraction. Account Abstraction (ERC-4337) enables sponsored transactions via paymasters, used by apps like CyberConnect. Networks like Polygon use a meta-transaction relay system. Starknet and zkSync employ fee delegation at the protocol level. Each method has distinct trust and finality trade-offs for the sponsoring entity.

Sequencer profit is the hidden subsidy. L2 sequencers, operated by entities like Offchain Labs or the OP Collective, profit from ordering transactions and capturing MEV. A portion of this sequencer revenue funds the gas abstraction pool. This creates a circular economy where user growth directly finances the subsidy, a model perfected by Arbitrum's sustained dominance.

The endgame is multi-chain gas markets. Protocols like Biconomy and Pimlico are building unified paymaster networks that let dApps sponsor gas across any EVM chain. This abstracts not just cost, but complexity, moving towards a future where users never hold native gas tokens. The competition shifts from cheap gas to the best developer subsidy program.

Sequencer cartels are inevitable. The economic model of rollups incentivizes centralization for profit. A dominant player like Arbitrum's Offchain Labs captures fees from all transactions, creating a natural monopoly that is difficult to dislodge due to network effects and liquidity.

Fragmentation is a tax. Users don't pay for L2 gas; they pay for the bridging and sequencing overhead. Every hop between chains like Optimism and zkSync adds latency and fees, with protocols like Across and LayerZero acting as toll collectors on this new financial plumbing.

The abstraction layer extracts rent. Solutions like UniswapX and intent-based architectures promise a seamless experience but merely obfuscate the cost. The fee is still paid, but now a new intermediary (e.g., a solver network) captures value, shifting rent extraction from the chain to the application layer.

Evidence: Arbitrum and Optimism sequencers generate over $1M in daily profit from MEV and fees. This revenue is a direct transfer from users and dApps to a single corporate entity, validating the rent extraction thesis.

Gas sponsorship is table stakes. Protocols like Pimlico and Biconomy abstract gas for users, shifting the cost to dApps. This creates a two-sided market where L2s compete for dApp subsidies and users chase the best UX.

The battle is for the payer. L2s like Base and Arbitrum will offer native gas credit programs to attract high-volume dApps. The winning model will be the one that optimizes for developer retention, not just user acquisition.

Account abstraction enables monetization. With ERC-4337 and EIP-7702, wallets become fee markets. L2s can implement dynamic fee routing, taking a cut from sponsored transactions, turning UX into a direct revenue stream.

Evidence: Coinbase's Base already subsidizes gas for its users, a strategy that directly fueled its initial growth. Arbitrum's Stylus and zkSync's Boojum upgrades are fundamentally about reducing the underlying compute cost that these programs must cover.