Ethereum's Layer 2 networks processed more daily transactions than mainnet throughout 2026, with combined L2 throughput reaching approximately 1,270 user operations per second versus 20.4 on L1. The scaling strategy worked. What did not work, according to on-chain data, is value accrual back to t...
"L2s are parasitic when they take more priority transactions from the base layer than they add." — Anatoly Yakovenko, Co-founder, Solana Labs
Ethereum's Layer 2 networks processed more daily transactions than mainnet throughout 2026, with combined L2 throughput reaching approximately 1,270 user operations per second versus 20.4 on L1. The scaling strategy worked. What did not work, according to on-chain data, is value accrual back to the base layer.
On-chain analyst @Tanaka_L2 published a breakdown on July 31, 2026 showing Ethereum L1 captured $88.4 million in Real Economic Value during Q2 — 4.9% of the $1.79 billion flowing through applications built on top of it. Standard Chartered estimated that Coinbase's Base chain alone removed approximately $50 billion from ETH's implied market capitalization by diverting transaction fees from mainnet. Daily L1 fee revenue that once exceeded $30 million now hovers near $500,000. Daily ETH burns fell to roughly 100 ETH, flipping the network from deflationary back to mildly inflationary at an annualized rate of approximately 0.23%.
ETH traded at $2,521 on September 13, 2026 — approximately 49% below its all-time high of $4,950 set in August 2025. Year-to-date, ETH declined roughly 32% versus Bitcoin's 11% drop, widening a performance gap that analysts attribute to structural fee leakage rather than cyclical weakness.
The Dencun upgrade, activated on March 13, 2024, introduced EIP-4844 — a separate data channel called "blobs" for Layer 2 rollups to post compressed transaction data to Ethereum. The upgrade reduced L2 data-posting costs by 90–99%.
Before Dencun, rollups posted data as calldata, paying standard gas prices. L2 data costs were the largest single fee category on Ethereum L1, generating tens of millions monthly. After Dencun, rollups shifted to blob transactions. According to Spark Research, in the first year post-Dencun, blob transactions paid roughly 1,020 ETH in EIP-4844 base fees plus approximately 2,000 ETH in EIP-1559 fees — totaling around $8 million. For context, L2s were paying $34 million per month in calldata fees before the upgrade.
Cumulatively since Dencun, Ethereum has earned approximately $26 million in blob fees — a fraction of prior calldata revenue. Token Terminal data shows monthly protocol revenue fell 60–80% in the quarters immediately following activation compared to equivalent activity periods in 2023.
The Pectra upgrade (May 2025) doubled the blob target from 3 to 6 per block. The Fusaka upgrade (December 2025) and subsequent BPO (Blob Parameter Only) forks raised the target further to 14 and the maximum to 21. Four capacity increases have occurred since Dencun; a fifth is under consideration.
Three L2 networks dominate the Ethereum ecosystem by revenue and total value locked (TVL) as of mid-2026:
| Network | TVL (Mid-2026) | Avg. Daily Revenue | Typical Tx Fee | Key Characteristic | |---------|---------------|-------------------|----------------|-------------------| | Arbitrum One | ~$17B | ~$55,025 | $0.01–0.05 | DeFi depth, institutional rails | | Base (Coinbase) | ~$10–11B | ~$185,291 | $0.01–0.03 | Corporate distribution, no token | | Optimism | ~$4B | N/A | $0.02–0.08 | OP Stack / Superchain licensing |
Base generated approximately $60–70 million in sequencer revenue in H1 2026, according to Coinbase SEC filings. As of January 2026, Base captured 70% of L2 fee revenue with $147,000 in daily earnings. Priority fees comprised 86.1% of Base's revenue. Base leads on throughput at 89 TPS and 382,500 daily active users.
The margin structure is notable. Base earned over $94 million in profit while contributing just $4.9 million back to Ethereum in blob fees — a ratio of roughly 19:1. Coinbase captures sequencer margins as corporate revenue with no token dilution. This is the clearest example of the L1-to-L2 value extraction dynamic.
Arbitrum's economics skew toward DeFi power users: perpetual futures, leveraged yield, and fixed-yield products. Optimism's strategic value in 2026 is defined more by the OP Stack — its open-source L2 framework powering the Superchain — than by its own mainnet transaction volume.
The EIP-1559 burn mechanism, activated in August 2021, was designed to remove ETH from circulation with every transaction. It worked: approximately 4.6 million ETH has been permanently destroyed since launch, according to Ultrasound.money.
However, the mechanism's efficacy depends on mainnet fee volume. After the Pectra upgrade in May 2025, the average daily burn fell to 3.26 ETH — down 71% year over year. Gas prices reached historic lows around 0.067 Gwei. The "ultrasound money" thesis — that ETH would be persistently deflationary post-Merge — requires average fees above approximately 16 Gwei to hold. Current levels are nowhere near that threshold.
With validator issuance running at approximately 1,700 ETH per day and daily burns far below that level, Ethereum's supply is growing. Binance data places the annualized inflation rate at roughly 0.23%. This compares favorably to Bitcoin's post-halving issuance of approximately 1.6%, but it is not the deflationary asset that was marketed to investors.
Approximately 42.4 million ETH is staked — 31.71% of circulating supply as of June 2026, per Dune Analytics. The staking lock-up provides some supply constraint, but it does not address the revenue problem.
Since September 3, 2026, Ethereum set a new record for blob usage: 6.7 blobs per block on a daily average, according to CoinTribune. This exceeds previous peaks recorded in late 2025 and indicates growing L2 adoption.
However, the network operates at only 40–50% of its current target of 14 blobs per block. The Fusaka upgrade and BPO forks expanded capacity 2.3x beyond pre-upgrade levels. Supply continues to outpace demand. The result: record utilization in absolute terms produces minimal fee pressure because capacity headroom remains substantial.
This dynamic illustrates the central paradox. More L2 usage drives more blob consumption, which should generate more L1 revenue. But Ethereum has expanded blob capacity faster than demand has grown, keeping blob fees near zero for extended periods. The base layer subsidizes the very networks extracting value from it.
Ethereum faces competition not only from its own L2s but from external data availability providers that offer rollups an exit from Ethereum's blob fees entirely.
Celestia, the first purpose-built DA blockchain, commands roughly 50% market share in the third-party DA sector. It has processed over 160 GB of rollup data. When Ethereum L2s paid $3.83 per megabyte to post data using blobs, Eclipse was paying Celestia $0.07 for the same megabyte — 55 times cheaper, according to BlockEden research.
EigenDA, built on EigenLayer's restaking infrastructure, targets approximately 100 MB/s throughput in its V2 release. For high-volume rollups posting 100 MB per day, EigenDA costs approximately $730 annually versus Celestia's $12,775, according to DA Layers analysis. Mantle uses a hybrid EigenDA-based approach.
Every major rollup framework — Arbitrum Orbit, OP Stack, Polygon CDK — supports Celestia as a DA option. Rollups using alternative DA are technically validiums rather than true rollups; they rely on external committees for data availability rather than Ethereum. The security trade-off is real but has not prevented adoption. Data availability represents 90% of L2 operating costs post-EIP-4844, making DA provider selection an economics-driven decision.
Ethereum core developers recognized the pricing failure. EIP-7918, included in the Fusaka upgrade (December 3, 2025), introduces a blob reserve price — a floor tied to execution costs.
Before EIP-7918, the blob fee auction could spiral to 1 wei during low demand, producing no meaningful revenue. The proposal introduces a constant called BLOB_BASE_COST, constituting a minimum price for blob gas expressed in execution-layer gas units. When the reserve exceeds the nominal blob base fee, the adjustment algorithm treats the block as over-target and stops pushing fees down.
According to Fidelity Digital Assets research, EIP-7918 represents a "direct attempt at displaying pricing power in the data availability market." Combined with PeerDAS (EIP-7594), which enables validators to sample blob data rather than downloading every blob in full, the upgrade aims to scale capacity while maintaining a revenue floor.
The market is still assessing whether EIP-7918 produces measurable improvement. Early indications suggest the floor prevents the worst-case zero-revenue scenarios but does not significantly increase average blob revenue during periods of moderate demand. The structural surplus of blob capacity remains the dominant price-setting factor.
Ethereum's fee revenue decline has coincided with competitive gains by alternative L1s. Ethereum's on-chain revenue dropped to fourth place in Q1 2026, trailing Solana, Tron, and BNB Chain, according to Standard Chartered research.
Solana captured 58% of decentralized exchange volume versus Ethereum plus its L2s at 40% — a reversal from prior years. Solana's 400ms finality and near-zero fees made it the default for memecoin launches, high-frequency DeFi, and consumer applications. Average daily non-vote transactions on Solana reached an all-time high of 112.6 million in Q1 2026, up 50% quarter over quarter, per Messari.
However, Solana's fee generation remains modest relative to transaction volume. Ethereum's all-time cumulative fee generation is $13.12 billion — roughly 22 times Solana's. The difference reflects transaction size: Ethereum mainnet still handles larger-value DeFi transactions and institutional flows. March 2026 data showed Gnosis, MetaMask, and Uniswap burning approximately 2,277 ETH ($3.63 million) in a single week — reflecting high-value settlement activity that prefers L1 security.
Ethereum's Q2 2026 on-chain yield dropped to a record low of 2.68%, per the DeFi Report. Competition for mainnet block space decreased as user activity migrated to L2s and competing chains, reducing priority fees and MEV capture.
Three conditions for ETH price recovery have been identified by analysts:
Blob scarcity. If RWA settlement volumes and stablecoin turnover grow enough to push blob demand past supply, fee pressure returns to L1 and the burn reactivates. Current utilization sits at 40–50% of target capacity. The gap is wide.
Active asset turnover. Stablecoins and tokenized real-world assets need to generate transaction velocity — not merely sit idle on-chain. USDC and USDT supply on Ethereum exceeds $100 billion, but settlement frequency remains below levels that would generate meaningful L1 fees.
ETH as reserve asset. Institutions holding ETH for settlement collateral rather than merely using the network would create sustained demand independent of fee economics. Corporate treasuries like BitMine are absorbing supply from spot markets, but this remains marginal.
The Glamsterdam upgrade, Ethereum's largest protocol change since The Merge targeting 10,000 TPS and 78.6% lower gas fees, was delayed from June to Q3 2026. Whether additional capacity expansion helps or further depresses fees depends entirely on whether demand growth outpaces supply.
Ethereum's scaling roadmap achieved its stated objective: L2 transaction costs dropped 90–99%, and user activity migrated off mainnet. The unintended consequence is that the economic value of that activity migrated with it. L2 sequencers — particularly Coinbase's Base — retain the vast majority of fee revenue while paying Ethereum a fraction in blob costs.
The value accrual gap is not a temporary condition. It is the direct result of architectural decisions: blob pricing that undercuts the base layer's own fee market, capacity expansion that outpaces demand growth, and the absence of a mechanism that ties L2 economic activity back to ETH as an asset. EIP-7918 introduced a floor, not a solution.
Whether this matters for Ethereum's long-term security budget depends on whether blob demand eventually saturates expanded capacity. Current data does not support that trajectory. The base layer has become a settlement utility with declining revenue — a viable infrastructure service, but one whose native asset lacks a clear path to capturing the economic growth occurring above it.