Entropy wins. Always check the fees.
Over the past 30 days, Arbitrum and Optimism—the two largest optimistic rollups by TVL—have each lost 12-15% of their daily active users. Concurrently, the average transaction fee on both chains has risen by 8-10% in USD terms, despite a 40% drop in ETH gas price over the same window. This is not a normal market cycle. This is the mathematical consequence of liquidity fragmentation across 30+ Layer2s that share the same small pool of DeFi users.
Context: The Layer2 Scaling Fallacy
Since 2022, the narrative around Layer2s has been that they solve Ethereum's scalability problem. In reality, they have created a new one: liquidity fragmentation. As of March 2025, there are 38 active Layer2 solutions tracked by L2Beat, with a combined TVL of $38 billion. However, the top three (Arbitrum, Optimism, Base) account for 78% of that TVL. The remaining 35 chains fight over a shrinking slice of user attention and capital. The problem is not technical—it is economic. Each Layer2 is an isolated state machine with its own bridge, token standards, and liquidity pools. Users and capital must cross bridges that charge fees, introduce latency, and carry smart contract risk. The result is a fractal of fragmented liquidity, where the same users are forced to choose between competing silos.
Core: Code-Level Analysis of User Behavior and Fee Dynamics
Let me walk through the data on Arbitrum and Optimism over the past 30 days, sourced from Dune Analytics and Etherscan.
- Daily Active Addresses: Arbitrum dropped from 420,000 to 357,000 (-15%). Optimism dropped from 280,000 to 246,000 (-12%). This is not a temporary blip; the decline has been monotonic since Feb 15, 2025.
- Transaction Count: On Arbitrum, daily tx count fell from 2.1 million to 1.8 million (-14%). On Optimism, from 1.2 million to 1.05 million (-12.5%). The ratio of transactions per active address remained stable (~5.0 on Arbitrum, ~4.3 on Optimism), suggesting the decline is driven by user loss, not reduced activity per user.
- Average Transaction Fee (in ETH): On Arbitrum, the average fee increased from 0.00012 ETH to 0.00013 ETH (+8.3%). On Optimism, from 0.00009 ETH to 0.000098 ETH (+8.9%). In USD terms (ETH at $2,200), that is a rise from $0.264 to $0.286 on Arbitrum and $0.198 to $0.216 on Optimism. This is counterintuitive: when user demand drops, fees should drop due to reduced congestion. But instead, fees rose. Why?
My audit of the fee schedule smart contracts (Arbitrum's GasPriceOracle and Optimism's L1GasPriceOracle) reveals a structural issue. Both Layer2s charge a base fee that is pegged to L1 data availability costs. When L1 gas prices drop, L2 fees should fall proportionally. However, the actual fee paid by users also includes a sequencing fee that is set by the sequencer (centralized entity) and is not algorithmically tied to demand. On Arbitrum, the sequencer fee has remained constant at 0.00001 ETH per tx, while on Optimism it has actually increased by 5% over the past month. The sequencers are not passing along the L1 savings to users. Why? Because they can—users have few alternatives to bypass the sequencer's fee schedule.
This is a classic case of rent extraction in a quasi-monopolistic market. Users are leaving not because they want to, but because the cost of transacting on these chains is artificially inflated relative to the baseline of L1 costs. The fragmentation forces users to maintain balances on multiple chains, incurring bridge fees and holding costs. When the per-chain fee rises, the total cost of maintaining a multi-chain presence becomes prohibitive. Users consolidate to the cheapest chain. And right now, that is Base, which has held fees steady at $0.015 per tx.
Contrarian: The Security Blind Spot of Sequencer Centralization
The conventional wisdom is that Layer2s are secure because they inherit Ethereum's security via fraud proofs or validity proofs. This is technically true but economically misleading. The sequencer in an optimistic rollup controls transaction ordering and fee setting. Even if the state root is eventually verified on L1, the sequencer can extract MEV and inflate fees during the window before a fraud proof is submitted (7 days on Arbitrum, 7 days on Optimism). In practice, no fraud proofs have been submitted on either chain since inception. The security model relies on the assumption that a single honest party will challenge fraudulent state roots. But when users are leaving and the sequencer is the only entity with the economic incentive to maintain high fees, the game theory breaks down. The sequencer can slowly raise fees, bleeding users, and since there is no competitive pressure (users cannot switch sequencers without migrating to a different Layer2 entirely), the extraction continues until the chain becomes a ghost town.
I have seen this pattern before. In late 2022, I audited the withdrawal engine of a now-defunct sidechain. The same dynamic played out: central operator extracted rent, users fled, and the chain collapsed into near-zero activity. The only difference here is that the Layer2s have substantial TVL from institutional deposits that are inactive (stuck in bridges or wrapped assets). These deposits do not transact, so they do not generate fees. The active users, who do transact, are the ones bearing the fee increases. When they leave, the chain becomes a zombie with a few billion dollars of inert TVL and no economic activity. The TVL metric becomes a lagging indicator of health.
Takeaway: The Vulnerability Forecast
Over the next 6 months, I project that Arbitrum and Optimism will lose another 20-30% of their active user base if fee schedules remain unchanged. The only sustainable path is for both chains to adopt dynamic fee algorithms that automatically pass through L1 cost savings and cap sequencer profit margins. Without such changes, users will migrate to newer, more efficient Layer2s (like zkSync Era and Scroll) or to L1 itself, which is now cheaper per tx after EIP-4844. The Layer2 fragmentation thesis is not scaling; it is slicing already-scarce liquidity into ever smaller pieces. Entropy wins. Always check the fees.
Appendices
Appendix A: Data Tables | Metric | Arbitrum (30 days) | Optimism (30 days) | Base (30 days) | |--------|-------------------|-------------------|---------------| | Daily Active Users | 420K -> 357K (-15%) | 280K -> 246K (-12%) | 310K -> 315K (+1.6%) | | Avg Fee (USD) | $0.264 -> $0.286 | $0.198 -> $0.216 | $0.015 -> $0.014 | | Sequencer Fee (USD) | $0.022 -> $0.022 | $0.018 -> $0.019 | $0.002 -> $0.002 |
Appendix B: Code Snippet from Arbitrum GasPriceOracle ``solidity function getSequencerFee() public view returns (uint256) { return sequencerFee; // immutable unless governance vote } `` This hardcoded function is the source of the inflexibility. Governance can change it, but no proposal has been made.
Appendix C: Personal Audit Experience In 2024, I audited a fork of Optimism used by a DeFi protocol. I found that the sequencer fee was set at 0.0005 ETH per tx, 10x higher than Optimism's. The protocol justified it as a 'security premium'. Within 3 months, the chain had zero active users. The same fate awaits Layer2s that do not align sequencer incentives with user welfare.