Skip to content
Crypto.world

Research

Transaction Costs and Speed in the Ethereum Ecosystem: Scalability of the Mainnet and Layer 2s

Ethereum has historically been slower and more expensive than competing chains like Solana, and its scaling strategy now depends on two levers: raising the Mainnet's block gas limit and expanding…

Transaction Costs and Speed in the Ethereum Ecosystem: Scalability of the Mainnet and Layer 2s
arXiv
2606.22206
Published
2026-06-20
Authors
Meghan Ambrosia, Bruce Mizrach

AI summary

Overview

Research area: Blockchain market microstructure and scaling economics — a cross-chain empirical comparison of transaction speed (TPS) and transaction fees across Ethereum Mainnet, its Layer 2 rollups, Solana, and Polygon.

Technical level: Intermediate. The paper uses standard time-trend and OLS regressions with HAC standard errors, but the concepts (gas limits, blobs, rollups) are explained from first principles, so a reader with basic blockchain familiarity can follow it.

Scope (one sentence): The paper documents how Ethereum's 2024–2026 protocol upgrades (Dencun, Pectra, Fusaka, and two blob-parameter-only forks) changed transaction throughput and fees on the Mainnet and its L2s relative to Solana and Polygon, and projects when Ethereum's ecosystem will catch up.

What This Paper Is About

Ethereum has historically been slower and more expensive than competing chains like Solana, and its scaling strategy now depends on two levers: raising the Mainnet's block gas limit and expanding "blobs," the cheap data space L2 rollups use to post transactions back to the Mainnet. The authors measure whether these upgrades actually improved speed and cost between January 2024 and March 2026, estimate the quantitative relationship between gas limits/blobs and outcomes, and then extrapolate those relationships to forecast when Ethereum and its L2s will match or beat Solana on TPS and fees.

Key Contributions

  1. Joint coverage of five protocol changes. The paper documents the combined effect of Dencun (March 2024), Pectra (May 2025), Fusaka (December 2025), and the two blob-parameter-only forks BPO-1 (December 2025) and BPO-2 (January 2026), together with the gas limit increases, on Mainnet transaction speed and fees. The authors state that no prior work covers all of these upgrades together.

  2. A four-way cross-chain comparison. Rather than comparing Ethereum to a single rival, the paper jointly studies the Mainnet, 20 ETH-fee-denominated L2 networks, Solana, and Polygon over the same window using a common TPS and median-fee methodology.

  3. Estimated structural links from protocol parameters to outcomes. The authors estimate how much Mainnet TPS rises per unit of gas limit and how much L2 TPS and L2 fees respond to each additional target blob.

  4. Quantitative TPS and fee forecasts, including a policy counterfactual. They project convergence dates under the current trend and separately under EIP-7938 from the February 2026 L1 Strawmap, which proposes a ten-fold gas limit increase every two years for four years.

Main Findings

  • Mainnet speed roughly doubled but stayed small in absolute terms. Average daily Mainnet TPS rose from 13.67 in 2024 Q1 to 25.78 in 2026 Q1 (annual averages: 13.59 in 2024, 16.61 in 2025).

  • L2 throughput nearly tripled. Combined L2 TPS ("L2 Total") rose from 78.60 in 2024 Q1 to 226.92 in 2026 Q1, with annual averages of 124.00 in 2024 and 227.08 in 2025.

  • Solana remained the fastest chain in the sample. Solana TPS rose from 804.72 in 2024 Q1 to 1,303.46 in 2026 Q1 (819.28 in 2024, 1,054.77 in 2025). Polygon rose from 45.72 to 91.51 TPS over the same quarters.

  • Gas limits are the main driver of Mainnet speed. A 10,000,000-unit increase in the gas limit is associated with 3.52 additional TPS. The median daily gas limit rose from 30 million in January 2024 to 60 million by December 2025. At roughly 60 million gas, the model implies 24 TPS, close to the March 2026 average of 25.59.

  • Baseline projection keeps the Mainnet far below 100 TPS. Under the estimated gas limit trend (a constant of 2.78 × 10^7 and a daily trend of 7.44 × 10^4, i.e. about 2.2 million units per month, adjusted R² of 0.88), the Mainnet is not projected to reach 100 TPS until February 2034 — described in the paper as less than one-tenth of Solana's current speed.

  • EIP-7938 pulls the 100 TPS date forward by six years. Applying the proposed hundred-fold gas limit increase beginning in 2027, the authors predict the Mainnet reaches 100 TPS by January 2028.

  • Blobs drive L2 speed. Each additional target blob is associated with 10.75 additional L2 Total TPS. With continued blob expansion at the BPO increment rate, L2 Total TPS is projected to surpass Solana's predicted TPS in March 2029, converging at 1,820 TPS.

  • Mainnet fees collapsed by more than 99%. The median Mainnet fee fell from $3.786300 in 2024 Q1 to $0.012283 in 2026 Q1, a decline of 99.68% (annual medians: $1.736066 in 2024, $0.164527 in 2025).

  • L2 leader fees fell by more than 99% as well. The "L2 Leaders" average (Arbitrum, Base, and Optimism) fell from $0.180219 in 2024 Q1 to $0.001512 in 2026 Q1, a decline of 99.16%.

  • Solana's fees fell much more slowly, and Polygon's rose. Solana's median fee went from $0.000687 to $0.000496, a decline of 27.80%. Polygon's median fee rose 19.62%, from $0.005742 to $0.006868, driven by a first-quarter 2026 surge in which daily total Polygon fees surpassed Ethereum's for the first time in February.

  • Gas limits also drive Mainnet fees. Every 5,000,000-unit increase in the gas limit is associated with a 59.2% decline in median transaction fees. Mainnet median fees are projected to converge with Solana's in August 2027.

  • Blobs drive L2 fees. Each additional target blob is associated with a 4.3% decline in median L2 leader fees. Blob-by-blob, the L2 leader fee fell from $0.2107 before blobs, to $0.0095 after Dencun's three target blobs (a 95.5% reduction), to $0.0019 at six blobs under Pectra, to $0.0013 at 10 blobs under BPO-1, and edged up to $0.0015 at 14 blobs under BPO-2. The introduction states that L2 fees are projected to converge with Solana's predicted fee in October 2027 (the abstract states October 2026 for lower median L2 fees).

  • Optimism is the cheapest L2 in the sample. Since May 2025, Optimism's median fees have averaged 81.7% below Solana's over eleven months; it is on average 78.1% below Arbitrum and 75.2% below Base.

  • Blob space remains heavily underused. Average blob usage tracked around one to two blobs per block throughout the sample, well below both target and maximum, and the share of blocks exceeding the blob target fell to near zero from mid-2025 onward.

  • Gas usage, by contrast, routinely exceeds the target. Between 37.7% and 48.5% of blocks exceeded the gas target on any given day, with no clear downward trend, even as the limit rose.

  • Rollup architectures diverged sharply. Optimistic rollup TPS rose from 29.67 in January 2024 to 195.22 in March 2026, while ZK-rollup TPS peaked at 33.57 in March 2024 and fell to about 7.82 by March 2026.

  • Solana's headline throughput is inflated by voting. Vote transactions averaged 71.5% of Solana's total transactions in the sample, which is why the authors benchmark against non-vote transactions.

  • Solana dominates DEX volume. In 2026 Q1, Solana accounted for 58% of DEX USD volume, Ethereum 22%, Base 12%, other L2s 6%, and Polygon 1% — meaning Ethereum plus its L2s combined reached 40%.

  • The L2 ecosystem now exceeds the Mainnet in activity. Total L2 transactions grew from 135.6 million in January 2024 to 543.8 million in March 2026, and Arbitrum and Base each individually exceeded Mainnet monthly transaction counts for most of the sample.

Methodology in Plain English

The authors assemble a daily panel from January 2024 through March 2026: full Ethereum transaction history from Google BigQuery, a public Dune query for the 20 ETH-fee L2s, another Dune query for Solana, and BigQuery's Polygon dataset. Fees quoted in native tokens are converted to USD using CoinGecko daily closing prices. TPS is computed as the sum of daily transaction counts divided by the number of seconds in a day, then aggregated to monthly and quarterly averages.

To explain what drives outcomes, they run two simple regressions. First, they regress the median daily block gas limit on a linear time trend from January 2025 onward, which gives them a forecastable path for the gas limit. Second, they regress Mainnet TPS on the gas limit over the same window. For the L2s, they regress total L2 TPS and median L2 leader fees on the target blob count. All regressions use heteroskedasticity- and autocorrelation-consistent standard errors with a Bartlett kernel. Forecasts are produced by extrapolating the time trend for Solana and combining the projected gas limit path (or blob path) with the estimated coefficients for Ethereum and the L2s. A separate scenario swaps in the EIP-7938 gas limit schedule.

Why This Matters

The paper turns a popular narrative — "Ethereum is scaling" — into measured parameters: how many TPS a unit of gas limit buys, how much a blob buys, and how many dollars of fee a 5-million gas limit increase removes. For research, it provides an empirically estimated bridge between protocol parameters and market outcomes, which is more useful than simple before-and-after comparisons around hard forks.

Real-world applications:

  • Protocol governance and upgrade planning. The EIP-7938 counterfactual (100 TPS by January 2028 versus February 2034 under the trend) gives developers a directly comparable cost-benefit picture for aggressive gas limit schedules.
  • L2 economics and rollup operator decisions. The 10.75 TPS per target blob and 4.3% fee reduction per blob estimates inform how much additional blob capacity L2s should plan for.
  • Exchange and market-maker venue selection. Given that Solana holds 58% of DEX volume while Ethereum plus L2s hold 40%, and that fee levels differ by orders of magnitude, the fee and speed series are directly relevant to routing and settlement decisions.
  • Fee forecasting for users and dApp operators. The projected convergence dates (August 2027 for Mainnet versus Solana fees; March 2029 for L2 TPS versus Solana) give a timeline for when cost-based chain selection should change.

Industry relevance: the findings show that Ethereum's scaling strategy has worked on cost — Mainnet fees down 99.68% — but that throughput remains an order of magnitude below Solana even after the upgrades, so competitive positioning still depends on the L2 layer as much as the base layer.

Future Directions

  • Will Blob Usage Catch Up to Capacity? Average blob usage stayed at one to two blobs per block while the target rose to 14, and the share of blocks exceeding the blob target fell to near zero from mid-2025 onward. The natural question is whether L2 demand will eventually fill that space, which determines whether the projected March 2029 convergence with Solana is realized.

  • Does EIP-7938 Get Adopted, and At What Cost? The paper models a ten-fold gas limit increase every two years for four years, but it does not report measurements of the resulting node hardware burden or centralization risk. Whether the Strawmap's one-gigagas-per-second target is achieved without degrading decentralization is left open.

  • How Do Vote-Transaction Reforms Change the Benchmark? Solana's Alpenglow upgrade, expected in 2026 Q3, will move vote transactions off-chain. The paper benchmarks using non-vote transactions, but whether the comparison shifts once votes leave the block space is untested.

  • Do Convergence Forecasts Survive Structural Breaks? The forecasts assume the trend in gas limits and blob expansion continues. The paper's model inputs predate the expected Glamsterdam upgrade in 2026 Q3 and its 200 million gas limit floor, so re-estimating after those changes would test the stability of the August 2027 and March 2029 convergence dates.

Target Audience

Blockchain researchers and graduate students in market microstructure and financial economics; protocol developers and Ethereum Foundation contributors working on scaling roadmaps; L2 operators and rollup economics analysts; and quantitative traders or infrastructure teams who need empirical fee and throughput benchmarks when choosing settlement venues. The paper is also useful to regulators and industry analysts seeking data-grounded comparisons of blockchain network performance rather than marketing claims.

Authors’ abstract

We study the evolution of transaction speed and fees from January 2024 through March 2026, comparing Ethereum Mainnet and its Layer 2 (L2) networks, as well as Solana and Polygon. Ethereum has undergone upgrades that have increased block size and blob count. These upgrades have doubled transactions per second (TPS) on both the Mainnet and the L2 networks. Mainnet median fees have fallen from over \$2 to under \$0.02, and L2 median fees have fallen more than 95% from \$0.05…

The full abstract on arXiv
Read the original paper