How much does an L2 pay to Ethereum? From 0.14% to 189.89%

Ethereum charges $0.2057 for its own transaction and 0.006869 cents for a Layer 2 network transaction: 2,995 times more for its own. In the 90 days between July-15 and Oct-12-2026, the 22 L2s with complete income and cost data (L2: chains that execute transactions on their own and publish the result to Ethereum for final settlement) billed $70,616,595.42 to their users in gas, the fee each network charges to execute an operation, and paid $195,891.89 to Ethereum to settle their batches. One dollar for every 360, or 0.2774%. This percentage is what a reader might expect as an answer, yet it falls apart as soon as the 22 rows are opened: the median chain pays 12.2718% of what it bills—44.2 times the aggregate—and Manta pays Ethereum 189.8852% of its income. The aggregate ratio, the median, and the 2,995 factor are proprietary calculations based on the window totals. What follows is the full table of the 22 chains, the arithmetic reason why smaller chains pay a larger fraction, and the breakdown of the bill between blobs and ordinary gas.

The question "how much do L2s pay to Ethereum" admits a single number, yet that number describes none of the 22. The other twenty-one are spread between Polygon PoS's 0.1384% and Manta's 189.8852%, with five networks not spending a single cent on blob fees in the 90 days between July-15 and Oct-12-2026. The aggregate and the distribution answer different questions, and both are provided below with their respective table.

How much did the 22 L2s pay to Ethereum between July-15 and Oct-12-2026?

The 22 chains with both complete series billed $70,616,595.42 to their users and paid $195,891.89 to Ethereum in the 90 days between July-15 and Oct-12-2026. The ratio, a proprietary calculation based on the two published totals, results in 0.2774%: a one-dollar toll for every 360 billed.

The magnitude is better understood from the other side of the collection. In those same 90 days, Ethereum collected $38,234,889.05 in fees from all its users, including L2 batch transactions, so the $195,891.89 from the 22 L2s represents 0.5123% of what Layer 1 earned on its own—one dollar out of every 195, proprietary calculation. On the median day of the window, Ethereum received $1,833.21 from the 22 chains combined and $340,942.81 from its own activity.

Per unit of use, the contrast is even greater. The five marked rows in the table are derived by dividing the window totals and are CleanSky calculations, not figures published by growthepie.

L2s move fifteen times more transactions than the chain they settle on, yet what they leave behind is equivalent to half a percentage point of what Ethereum collects from its own users.

Why does the 0.2774% aggregate not describe any single chain?

Because the aggregate is a weighted average, and the weight is held by a single chain. Robinhood billed $50,799,996.71 of the $70,616,595.42 in the July-15 to Oct-12-2026 window, with its own ratio of 0.1419%, and Base billed another $9,916,816.72 at 0.3316%. Together, they provide the bulk of the denominator and pull the joint result toward their own level.

Removing them from the calculation measures how much of that figure is Robinhood. The three trimmings are proprietary calculations based on the per-chain series.

There is a second problem with that denominator: Robinhood's billing within the window is not a flat line that can be taken as a reference level. Of its $50.8 million, $40.79 million was concentrated in September 2026, and $28.69 million of that occurred in the first six days of the month. In the first twelve days of October 2026, it billed $704,406. The 0.2774% describes a window containing a specific spike.

We previously published a proprietary measurement for that chain with a different sample: the 0.06% we calculated for Robinhood Chain based on $28 million over 30 days as of Sept-7-2026, and the conclusion then was that the settlement bill had ceased to be an issue for that chain. With the current sample, that reading holds for the large players but collapses for the median chain, which pays 12.2718%, and for Manta, which pays 189.8852%. The previous conclusion applies to the subject it measured and cannot be extended to all 22 chains.

Did the toll drop between July and October 2026?

The monthly ratio, a proprietary calculation based on growthepie daily series, moves from 0.5329% in July to 0.1595% in September and returns to 1.0457% in the first twelve days of October 2026. This series measures the size of the denominator, not the price of the toll: billing for the 22 chains jumps from $12,744,259.24 in August to $48,972,853.82 in September, while the amount paid to Ethereum goes from $54,004.75 to $78,127.64—and in October, the movement reverses.

The column that remains steady is the last one. The contribution of the 22 L2s to Ethereum's fee revenue is 0.6078% in July, 0.5066% in August, 0.4881% in September, and 0.5078% in October 2026: half a percentage point month after month, while the daily billing of the 22 chains fluctuated from $265,405 in October to $1,632,428 in September, a factor of 6.15. If there is one figure from this window useful for tracking next month, it is this one, rather than the ratio against billing, which shifts with every activity spike from a single chain.

Which two fee series does growthepie publish and which L2s are excluded?

growthepie publishes two fee series per chain, differing by a factor of twenty. The first, labeled "Revenue" in its dashboard, is what users pay to the chain in gas, which goes into the operator's coffers: for Base, this is $9,916,816.72 in the 90 days between July-15 and Oct-12-2026. The second is what users pay to applications living on that chain—$196,303,457 on Base in the same window—money that goes to protocols and never passes through the sequencer. This piece divides by the first. A reader opening the dashboard with the second will see twenty times more billing.

On the cost side, growthepie also offers two fields, and the one used in this piece is the total: Layer 1 settlement plus blob fees. This matches the "rent paid" field to the cent for 19 of the 22 chains, but not for three—Celo, MegaETH, and Manta—where the difference is exactly their blob line: $1,033.31 for Celo, $1,358.60 for MegaETH (with an $8.53 deviation from its declared blob spend), and $5.58 for Manta. These three have data availability outside of Ethereum, so "rent paid" deducts the blob expenditure they nonetheless paid to Ethereum. Anyone taking that field as "what is paid to Ethereum" is leaving that money out.

The window is 90 days rather than 91 because growthepie provides 91 dates up to Oct-13-2026, but the billing and cost series end on the 12th. Counting the 13th would mean counting a day without financial data. Furthermore, four chains with activity are excluded from all ratios due to a lack of paired series:

• zkSync Era: publishes 90 days of billing ($51,704) but no cost data, so counting it as zero would give it a 0.00% toll.
• Fraxtal: publishes 90 days of cost ($1,496.21 paid to Ethereum) but no billing data, so no division is possible.
• Arbitrum Nova: 48 days of cost against 90 days of billing.
• Linea: missing two days of cost.

Fraxtal's $1,496.21 does count toward what Ethereum receives—$199,073.82 with the four excluded chains added—but does not enter into any percentage. The method is the same one used in the scorecard of the ten largest chains by five fundamentals, using the same source.

What does each of the 22 chains pay per dollar billed?

This is the full table, sorted by billing. Both money columns are data published by growthepie for the 90 days between July-15 and Oct-12-2026; the ratio column is the quotient between them, calculated by CleanSky. The data layer is the label from the growthepie master file timestamped Oct-13-2026.

Three rows deserve specific mention. The minimum is set by Polygon PoS at 0.1384%, achieved by publishing its data availability in calldata—the data field of an ordinary transaction—without touching blobs. The maximum is Manta at 189.8852%: it pays Ethereum nearly double what it bills, making it the only one of the 22 above 100%, carrying negative profit on 81 of the 90 days in the window, from $9.58 in the red on July-15-2026 to $22.45 on Oct-12. And Taiko, at 65.6428%, is not there due to high blob costs: it publishes about 8 blobs a day, and its bill is almost entirely calldata and execution.

Between the extremes are two cases we have already measured separately, which the table contextualizes. Base pays 0.3316% of what it bills, consistent with our report that Base charges dollars for sequencing and pays cents per blob; and OP Mainnet pays 5.2500% compared to Base's 0.3316%, a contrast worth noting when reading about Superchain revenue sharing after Base's exit. For exchange chains, the table adds figures to the comparison of which exchange L2 truly captures value: Robinhood 0.1419%, Base 0.3316%, and Ink 3.8632%.

Why does a small chain pay a larger fraction than Base or Robinhood?

The relationship between size and toll is monotonic and strong: the Spearman rank correlation coefficient between billing and ratio is −0.8012 across the 22 chains, a proprietary calculation based on the two series. Sorted by billing and split into four brackets, the breakdown—also a proprietary calculation—looks like this.

There is no commercial agreement behind this: the cost of publishing a batch on Ethereum is nearly fixed and is distributed among the transactions within it, so the top five paid 0.1805% while the bottom six paid 38.7302%.

This results in the contrast summarizing the July-15 to Oct-12-2026 window: the aggregate for the 22 L2s is 0.2774% of what they bill, the median per chain is 12.2718%, and the simple average of the 22 ratios is 27.2008%. The median is 44.2 times the aggregate. All three are proprietary calculations.

How much of the $195,891.89 received by Ethereum consists of blob fees?

Of the $195,891.89 that the 22 chains paid to Ethereum in the 90 days between July-15 and Oct-12-2026, blob fees account for $15,497.25. The remainder—$180,394.64—is the ordinary gas of the L1 transaction that publishes the batch.

The blob—the temporary data space enabled by the Dencun upgrade in March 2024 to allow rollups to publish cheaply—represents 7.9% of the bill. Five of the 22 chains spent nothing on blobs during those 90 days: Polygon PoS, Ronin, Plume, Metis, and Gravity. Furthermore, the data layer label does not predict this expenditure because it is not exclusive: Celo ($1,033.31), MegaETH ($1,367.13), Mantle ($74.85), and Manta ($5.58) have data availability outside of Ethereum—on EigenDA for the first three, on Celestia for the fourth—yet still paid for blobs in the window. These four rows are the reason not to read the data layer column as a cost table, and they add context to our findings on EigenDA vs. Celestia in data costs.

Blobscan is another provider, with a different indexer, and publishes blob fees per rollup in ETH. Its figures represent the full historical record since Dencun, from March-13-2024 to Oct-12-2026 (943 days), without being trimmed to the 90-day window of the previous ratios. The 67 rollups labeled by Blobscan have paid 3,697.49 ETH in blob fees in that full history, or $9.27 million at the ETH price of Oct-13-2026 ($2,507.16); there are another 1,310.51 ETH from issuers not labeled by Blobscan. The ranking of the largest historical payers—Base 1,820.95 ETH; Arbitrum 483.09; Taiko 386.32; OP Mainnet 291.70; World Chain 149.00—does not match the window, nor should it: the two sources measure different periods and different universes. They agree that Base is the leader by far, and among the chains labeled by Blobscan, they agree that Arbitrum is second. Robinhood, which in the window is the second-highest blob payer according to growthepie at $4,412.74, is not labeled in Blobscan, nor are MegaETH or Celo.

The contrast between the two windows illustrates the magnitude of the regime change. Base paid an average of 1.9310 ETH per day in blob fees throughout those 943 days of Blobscan's full history. In the 90 days between July-15 and Oct-12-2026, it paid $4,988.41 in blobs according to growthepie, which is 0.0221 ETH per day: 87 times less per day. Yet it publishes more data than before: the equivalent of 9,418 daily blobs in the window, calculated by us by dividing the blob bytes published by growthepie by the size of a single blob, compared to the 8,917 per day derived from counting the 8,408,585 blobs in Blobscan's history. The comparison between 1.9310 and 0.0221 ETH per day is a proprietary calculation based on both files, with the Oct-13-2026 price applied only to the current side.

Taiko is the inverse case and explains why its 65.6428% cannot be read as a blob bill: it paid 386.32 ETH in blobs throughout Blobscan's full history since Dencun, the third rollup on the list, yet publishes only about 8 blobs a day in the window. It has almost stopped using them, and what it paid in the window is calldata.

How does the 0.2774% fit with JPMorgan's account of rollup payments?

There is a figure in circulation pointing to a different scale that should be placed alongside, with its author and date. In its May 2026 note, which we covered on May-20-2026 when discussing JPMorgan's thesis on rollup payments to Layer 1, the bank estimated the aggregate payment from rollups to Ethereum at $113 million in 2024—41% of L2 operator revenue—and at $10 million in 2025, a 90% drop in one year. Our 90-day window, roughly annualized, yields $0.79 million per year and 0.2774% of billing. The annualization is a proprietary calculation, and scaling 90 days to a year is a crude exercise, not a projection.

The three figures trend in the same direction: $113 million in 2024, $10 million in 2025, and $0.79 million annualized from the 2026 window. What cannot be done is to treat them as the same measurement. JPMorgan covers all rollups, while this piece covers 22 chains with paired series; the bank measures calendar years, while here we measure 90 days with a Robinhood spike included; and the cost breakdown of the note is not public, so it may include concepts—proof verification, pre-Dencun calldata—that these series do not contain. The 0.2774% is not a corrected version of the 41%: they are two different calculations based on two different universes of chains and two different gas regimes, separated by the drop in blob prices visible via Blobscan from the data supply side. What this window adds to JPMorgan's series is another year in the same direction.

What is measured in the July-15 to Oct-12-2026 window and what is not?

It is measured that the 22 chains paid $195,891.89 to Ethereum and billed $70,616,595.42 to their users in those 90 days; that 7.9% of that bill consists of blobs; that the aggregate ratio is 0.2774% and the median per chain is 12.2718%; and that the contribution of L2s to Ethereum's fee revenue remained between 0.4881% and 0.6078% during the four months of the window. With these numbers, a reader can judge two authored readings: JPMorgan's, which attributes rollup payments to blobs when 92.1% of what they pay is ordinary gas, and our September reading on Robinhood Chain, which holds for large players but not for the median chain.

What these series do not reveal is the margin remaining for a chain after its operating costs. Furthermore, the window rolls daily: it loses the oldest day and gains the most recent, so any re-evaluation requires re-downloading the series. On Nov-12-2026, we will publish a verification of the only number that has been stable here: the half-percentage point contribution to Layer 1 revenue.

Sources and links: growthepie — Fundamentals API per chain (billing, L1 and blob costs, transactions; download Oct-13-2026) · growthepie — Chain master with data layer label (timestamp Oct-13-2026) · growthepie — L2 Economics Dashboard (revenue, costs, and profit per chain) · Blobscan — Overall blob stats per rollup since Dencun (reading Oct-13-2026) · Blobscan — Blob Statistics Dashboard · EIP-4844 — Blob transactions enabled by Dencun in March 2024 · CoinGecko — ETH price used to convert Blobscan ETH figures ($2,507.16, Oct-13-2026)