L2 fees fell another 38% — but the user moved to Base, not Arbitrum
Median L2 fees fell 38% over the last quarter. The user did not split evenly across rollups — Base captured the flow. Here is the gas math, the TVL share shift, and why Coinbase distribution mattered more than any sequencer upgrade.
Median Layer-2 transaction fees fell another 38% over the trailing quarter, according to the running aggregates on L2Beat’s activity dashboard. The median swap on Base now clears at around $0.018; on Arbitrum One, about $0.024; on OP Mainnet, roughly $0.031. Most of the drop is straightforward blob-gas math after the post-Pectra calibration. The interesting part is the user data. Daily active addresses, transaction count, and DEX volume have not split evenly across the three majors. Base captured the migration. Arbitrum, despite shipping Stylus, BoLD, and a meaningfully cheaper sequencer, ceded share. The reason is a distribution story, not a tech story.
What is at stake is the assumption — held across most of crypto Twitter and a non-trivial share of L1 governance — that rollups compete on technology. The last twelve months suggest that, at the margin, they compete on whose fiat onramp is one tap and whose stablecoin is native. Base has both, and the distance is showing up in the fee data, the TVL share, and the dApps that picked a side. Anyone planning Q3 deployments, pricing bridge flows, or routing strategies through our L2 gas dashboard should treat the Coinbase-distribution variable as the single most important input.
The actual fee numbers, by chain and operation
“L2 fees” is too coarse. The cost of a transaction on a rollup is the L1 data-posting cost (now blob gas) amortised across the batch, plus the L2 execution gas, plus whatever margin the sequencer takes. After EIP-4844 and the subsequent blob-target increases, the L1 component collapsed; the L2 execution component now dominates, and that is set by per-chain gas pricing. The table below pulls live medians from L2Fees.info and cross-checks against growthepie’s per-chain cost tracker.
| Chain | ETH transfer | ERC-20 transfer | Uniswap v3 swap | NFT mint (ERC-721) | QoQ change |
|---|---|---|---|---|---|
| Base | $0.004 | $0.009 | $0.018 | $0.024 | -41% |
| Arbitrum One | $0.005 | $0.011 | $0.024 | $0.030 | -35% |
| OP Mainnet | $0.007 | $0.014 | $0.031 | $0.038 | -37% |
| Linea | $0.006 | $0.013 | $0.028 | $0.034 | -39% |
| zkSync Era | $0.011 | $0.022 | $0.048 | $0.061 | -31% |
| Scroll | $0.012 | $0.024 | $0.052 | $0.066 | -29% |
| Blast | $0.005 | $0.012 | $0.025 | $0.032 | -36% |
The catch is that the absolute differences are small in cents but enormous as ratios. A swap on Base costs 25% less than the same swap on Arbitrum and 42% less than on OP Mainnet. For a user trading once a week, that gap rounds to noise. For a market maker rebalancing inventory or a points-farmer cycling positions 50 times a day, it is the difference between viable and not. The bots followed the cheapest venue first, retail followed where the bots created the best liquidity, and the flywheel handed Base a meaningful chunk of Arbitrum’s volume.
TVL share has shifted — Base now leads on every flow metric except total deposits
The headline TVL chart still puts Arbitrum One at the top of the L2 TVL stack at roughly $14bn, with Base at $11bn, OP Mainnet at $6bn, and the long tail (Blast, Linea, Scroll, Mantle, Mode, zkSync) accounting for the balance — figures from L2Beat’s TVL page. That snapshot misses the velocity story. Base now leads on:
- Daily active addresses — Base averages 1.4M-1.7M DAA, Arbitrum 650k-900k, OP 340k-480k, per growthepie’s fundamentals tab.
- Daily transactions — Base routinely settles 6M-9M tx/day; Arbitrum sits in the 2.5M-4M band; OP between 1M and 2M.
- DEX volume — Aerodrome alone has done $30bn-$40bn in trailing-30-day volume, more than every Arbitrum DEX combined on most days, per DeFiLlama’s per-chain DEX breakdown.
- Stablecoin transfer count — native USDC transfers on Base have crossed 60% of all L2 stablecoin transfer count in recent weeks.
Arbitrum still wins on locked capital — the long-tenured GMX, Gains Network, Pendle, and Camelot positions are sticky in ways that DEX volume is not. But the share of flow has moved decisively. The composition of users that touched each chain in the past month also tells the story: Base’s median address holds $110-$180 and transacts 14-22 times per month; Arbitrum’s median address holds $420-$600 and transacts 5-7 times. Base is winning the retail funnel; Arbitrum is keeping the size-weighted depositor.
Native USDC was the unlock — distribution mattered more than tech
The pivotal change for Base was Circle issuing native USDC on Base in September 2023, replacing the bridged USDC.e that every other L2 still relies on for the long tail. Native USDC means three things that matter in compounding ways: it is mintable and redeemable through Circle’s primary venues without bridging, it routes through CCTP for fast cross-chain settlement, and — most importantly — it is the asset Coinbase customers receive when they “Send to Base” from a Coinbase account at zero fee.
That last point is the entire story. Coinbase has ~110 million verified retail accounts. When the “Send to Base” option appears next to “Withdraw to wallet” with a fee of $0 and a confirmation time of seconds, the friction collapses. The equivalent for Arbitrum is a third-party bridge, a $2-$8 ETH withdrawal from Coinbase, then a deposit to a bridge contract. The user experience is not comparable; it is in a different category. Arbitrum’s official bridge UI is excellent, but it is still a separate app that the median user does not have, and that fact alone explains the daily-active-address gap.
The protocols that picked Base early — Aerodrome (the Velodrome fork purpose-built for Base liquidity), Friend.tech (which generated $50M+ in protocol revenue at peak before fading), and Farcaster (which built its entire payments stack on Base) — benefited from the same distribution. Aerodrome’s protocol stats show veAERO emissions paying out at APRs that simply could not be matched on chains where the underlying volume was thinner. Friend.tech’s fee model required a chain where 50,000 micro-transactions per day did not bankrupt the user; Base was the only credible option at launch. Farcaster’s Frames and tipping flows depend on sub-cent transactions; the engineering team would have built somewhere else if Base had charged Arbitrum prices.
What Arbitrum has shipped — and why it has not closed the gap
The frustrating part for the Arbitrum community is that the tech roadmap has, by most measures, executed better than Base’s. Stylus shipped to mainnet, enabling Rust and C++ contracts that run at fractional gas cost. BoLD — the permissionless dispute protocol — went live, removing the last centralised guardrail on the rollup’s security. The Nitro sequencer continues to be the most performant of the OP-Stack/Nitro family on raw throughput. The Arbitrum docs and the governance forum show a roadmap that any other rollup would happily trade for.
None of it closed the distribution gap. ARB token incentives via the STIP and LTIP programmes pushed billions in cumulative TVL but did not produce sticky retail engagement; the addresses that arrived for the incentives left when the incentives ended. The lesson is not that incentives don’t work — they obviously do for capital migration — but that they don’t substitute for a fiat onramp. Coinbase is, structurally, the most important fiat onramp in the West. Without an equivalent, Arbitrum’s user funnel has to be built one app at a time. That is a longer road, and the QoQ data shows it.
The OP Mainnet question — and the Superchain hedge
OP Mainnet sits in an awkward position: it is technically the parent stack that Base is built on, so every Base success technically validates the OP-Stack, but OP Mainnet itself has lost share to its own child chain. Optimism’s Superchain pitch reframes this as a feature — OP, Base, World Chain, Mode, Zora, Soneium, and Unichain are all OP-Stack chains, and the Superchain Interop upgrade (incrementally shipping through 2026) will allow native cross-chain messaging without bridges. If interop works as specified — sub-second cross-chain calls between any two OP-Stack chains — the competitive frame shifts from “Base vs OP” to “OP-Stack vs everything else.”
For now, that is a forward bet. OP Mainnet itself has the highest realised fee of the three majors in the table above, which is a function of less aggressive sequencer pricing rather than worse tech. The chain that will benefit most from Interop is the one that already has the users, and that chain is Base.
What the fee compression means for the next twelve months
A 38% QoQ decline in median fees is not a one-off. Blob-gas pricing is set by demand against a target of six blobs per block, with the cap raised to nine in the post-Pectra calibration. As more rollups batch and as compression improvements (notably the BLS-based attestation rollups proposed in EIP-7732) reduce per-tx data footprint, the trend continues. The implication is straightforward: by Q4, the median L2 swap will probably cost under a penny, and the differentiator between L2s will be UX and distribution, not gas.
- Watch native asset issuance. Tether’s native USDT, PayPal’s PYUSD, and any future bank-issued stablecoin will create the same flywheel Base got from USDC, on whichever chain wins them.
- Watch fiat onramp integrations. Robinhood’s Layer-2 (built on Arbitrum Orbit) is the most direct attempt to replicate the Coinbase-Base axis. Whether it ships with a real on-ramp button matters more than the chain’s gas price.
- Watch the Superchain Interop release. If cross-chain transfers between OP-Stack chains become as cheap as a single L2 transaction, the calculus changes for any app that needs to be on more than one chain.
The 38% fee drop is real and welcome. The user-flow story is the bigger one: rollups are no longer competing on tech, they are competing on whose stablecoin is one tap from a checking account. Base won the last round on that axis. Arbitrum and OP have the technical roadmap to compete on the next one — but only if they ship the distribution layer to match. Our market page tracks the per-chain DEX volume and stablecoin transfer trends, and the events calendar flags the upcoming sequencer upgrades and fee model changes. Both are the right places to track this from here.