The Early Days of the Ethereum Blob Fee Market and Lessons Learnt

TL;DR

This paper analyzes Ethereum's blob fee market introduced by EIP-4844, focusing on how blob data pricing affects rollups and on-chain data availability. It leverages transaction, mempool, and PBS-relay data from March–September 2024 to quantify block-packing inefficiencies and two peak-demand periods, finding up to a 70% fee loss from suboptimal packing. The study identifies a structural market-design flaw—subset bidding—and shows that current economic incentives for blob inclusion are weak, with most blob-related fees burned and only a small residual incentive for builders. It proposes concrete remedies, including multidimensional blob-priority pricing, gradual blob-target increases, faster price discovery, and improved packing plus flexible bidding mechanisms, to enhance efficiency, reduce delays, and better align incentives in a growing blob ecosystem. Overall, the work informs ongoing network upgrades by highlighting practical bottlenecks and actionable avenues to harness blob data pricing for scalable L2 data availability.

Abstract

Ethereum has adopted a rollup-centric roadmap to scale by making rollups (layer 2 scaling solutions) the primary method for handling transactions. The first significant step towards this goal was EIP-4844, which introduced blob transactions that are designed to meet the data availability needs of layer 2 protocols. This work constitutes the first rigorous and comprehensive empirical analysis of transaction- and mempool-level data since the institution of blobs on Ethereum on March 13, 2024. We perform a longitudinal study of the early days of the blob fee market analyzing the landscape and the behaviors of its participants. We identify and measure the inefficiencies arising out of suboptimal block packing, showing that at times it has resulted in up to 70% relative fee loss. We hone in and give further insight into two (congested) peak demand periods for blobs. Finally, we document a market design issue relating to subset bidding due to the inflexibility of the transaction structure on packing data as blobs and suggest possible ways to fix it. The latter market structure issue also applies more generally for any discrete objects included within transactions.

Figures (13)

• Figure 1: Blob usage by L2s. Figure \ref{['fig:adoption2']} shows the overall demand, while Figure \ref{['fig:shareOfBlobs']} visualizes the individual blob usage of the biggest L2s.
• Figure 2: Cumulative fees paid by blob transactions over time. We separate the EIP-4844 fee market (base fee) from the EIP-1559 fee market (base and priority fee).
• Figure 3: Development of various fee components during the spike in blob demand caused by LayerZero airdrop on 20 June 2024.
• Figure 4: Number of blobs submitted by the biggest nine L2s per type-3 transaction over time, each dot represents one type-3 transaction submitted by the respective L2. The grey vertical line indicates the spike in demand on 20 June 2024 related to the LayerZero airdrop.
• Figure 5: Distribution of priority fee (see Figure \ref{['fig:prio']}), gas usage (see Figure \ref{['fig:gas_used']}) and inclusion delay (see Figure \ref{['fig:inclusionDelayMs']}) of type-3 transactions submitted by the nine biggest L2s.