Scaling Ethereum: The Impact of EIP-4844 and the Blob Market

Introduction

The Ethereum ecosystem has come a long way in terms of scalability in recent years. One major milestone in Ethereum’s scalability roadmap was Proto-Danksharding (EIP-4844), which was included in the Dencun upgrade on March 13th. This upgrade introduced a new data structure (blobs) to enable rollups to batch transactions off-chain and post the proofs to Ethereum at a lower cost. Since the implementation of EIP-4844, the throughput and scalability of Ethereum and its associated rollups have climbed to over 220 TPS today and significantly reduced costs for rollups and their end users.

In this piece, we will examine EIP-4844 and Ethereum’s blob market in depth to understand its impact and how it will evolve.

Proto-Danksharding (EIP-4844)

EIP-4844 was an important upgrade that served as an intermediate step in Ethereum’s rollup-centric roadmap, laying the groundwork for highly scalable, cost-effective rollups. Its purpose was to improve the scalability of rollups on Ethereum and reduce their cost for end users. It achieves this by introducing blob-carrying transactions attached to blocks containing executed transaction data from rollups.

What are Blobs?

Blobs (Binary Large Objects) are large blocks of data (up to 128kb each) introduced as a part of EIP-4844 and designed to provide a cheaper way for rollups to add data to blocks. Blobs are a new transaction type included in each block with its dedicated gas market. The maximum size for blob storage per block is currently 0.75 megabytes.

Before EIP-4844, rollups would post batches of transaction information in L1 CALLDATA, incurring significant costs to the rollup, which would bear this cost and pass it on to their end users via higher gas costs.

Blobs are so much cheaper than CALLDATA because they are temporary, lasting only 18 days before being deleted. In addition, the EVM cannot access the data in these blobs; instead, the contents of the blobs are gossiped through a consensus client sidecar and not available to the execution client.

Since EIP-4844 was implemented on March 13th 2024, data fees from rollups have fallen dramatically from over $41m in March to just over $260k in August.

Blob Fee Market

Given blobs' unique role in the Ethereum network, EIP-4844 creates a new blob gas market that operates similarly to EIP-1559. As of EIP-4844, each block can contain up to 6 blobs.

If there are more than the target (currently 3) number of blobs in a block, the base fee will increase in the following block; if there are less than 3 blobs in a block, it will decrease, and if there are exactly 3, it will stay the same. While CALLDATA transactions are no longer used by rollups today, rollups have the additional option of utilizing CALLDATA, which may be cheaper depending on network conditions and demand for blobs. This can provide a floor on mainnet demand and a release valve for the blob market during high congestion.

Today, blobs still have considerable unutilised capacity, with over 40% of blocks containing 0 blobs and only 15% of blocks utilising their max capacity of 6 blobs per block. While this is still relatively low, it’s a considerable improvement from the low point in March, where over 80% of blocks contained 0 blobs. Interestingly, 6 blobs have remained the largest group behind 0 blobs over time, and blocks with 4 or 5 blobs have grown in occurrence over time. Another way of observing blob utilisation is to look at the average blob count per block over time.

Since the introduction of blobs in EIP-4844, average blob counts have remained below the target of 3 and currently sit at 2.1 blobs per block, illustrating the amount of unutilised capacity used in the network. While low, this is a considerable improvement from the low of 0.9 per block after the EIP-4844 upgrade in March.

Blob Fees

The blob base fee has remained very low on average, with only three weeks, when the average cost of blobspace rose above $0.01 per blob. The first was the surge of “blobscriptions” (inscriptions using blobs) in late March to early April, which caused blob space to rise due to demand for blob inscription transactions.

The second high blob fee period resulted from the LayerZero airdrop on June 20th, in which there was a surge in demand for transactions on the Arbitrum Network, which led to a surge in demand for blobspace. Blob base fees peaked at 7,471 gwei (around $3,450 per blob).

Total fees generated by Ethereum from blobs remain relatively low due to the persistent low utilisation and base fee. In September, the total daily fees generated by blobs remained below $50k, although at their peak in March, the total blob fees reached $234k.

While blobspace fees remain low, blob-carrying transactions (type-3 transactions) do pay gas fees for execution on the mainnet, with the average blob paying between $0.50 to $3 in execution fees.

Given that blob-carrying transactions aggregate blob transactions, the execution cost per blob comes down with the number of blobs included. Transactions carrying only one blob pay the highest fees, while those carrying 5 to 6 blobs incur much lower costs per blob.

Danksharding

Danksharding, which refers to the full realisation of rollup scaling, aims to reduce costs further for rollups and end users by increasing the space on Ethereum for rollups to post blob transactions. From a performance standpoint, this means reaching 100k TPS and enabling support for hundreds of individual rollups built on Ethereum. Danksharding aims to achieve this by expanding the number of blobs attached to blocks from the current 6 to 64 in Danksharding, an over 10x increase in the number of blobs per block. In addition, several changes would be made to how consensus clients operate to enable them to handle larger blobs.

Proposer Builder Separation

One dependency required to implement EIP-4844 was out of protocol proposer-builder separation (PBS). PBS separated the task of building and proposing blocks live on the mainnet in a limited form called “proto-PBS” through an optional out-of-protocol solution called MEV-Boost. Proto-PBS went live during the merge on the 15th of September, 2022. The MEV boost implementation is considered proto-PBS as it is not enshrined at the protocol level. It currently relies on centralised relays to facilitate the block production pipeline and enable access to the external block-building market.

PBS was a requirement for the Danksharding roadmap as it prevents solo stakers from generating expensive commitments and proofs for 32MB of blob data, which would require them to upgrade their hardware. Instead, specialised block builders would perform this computationally heavy work. In-protocol enshrined PBS (ePBS) is not a strict requirement for danksharding but a prerequisite for effectively implementing danksharding.

Data Availability Sampling

Data availability sampling (DAS) is another dependency required for the Danksharding roadmap. DAS enables validators to efficiently verify blob data to ensure it is available and correctly committed. Validators randomly sample data points and create proofs to ensure no validator has to check the entire blob. DAS is still in the research and development phase and is being deployed as a solution for light clients and future scaling solutions to efficiently verify data availability without downloading entire blocks.

Conclusions

EIP-4844 marks a significant milestone in the Ethereum scaling roadmap, introducing blobs as the first step in the danksharding roadmap. Blobs provide significantly lower costs for Ethereum rollups by creating a blob-specific transaction type and a dedicated fee market for blobs, reducing competition for blockspace for rollups while adding significant throughput to the scalability of Ethereum.

Today, blobs remain underutilised, with most blocks containing an average of only 2.1 blobs per block, below the target of 3 per block and considerably below the maximum of 6 per block. This has led to predominantly low blob base fees with a few notable brief exceptions. As blob demand increases and average blobs per block rise above 3, we should see a further reduction in 0 blob blocks and a more competitive blob fee market. This would result in higher blob fees for Ethereum and may cause some rollups to pass on these additional costs to end users through higher gas fees on rollups.

As Ethereum continues with the rest of the danksharding roadmap over the coming years, blobs per block will rise significantly from the current six blobs per block to 64. This is an over 10x increase in blob capacity, which will help further lower costs for rollups and increase Ethereum's throughput. In the roadmap to full danksharding are two prerequisite upgrades, including enshrined PBS (ePBS), which is not a strict requirement but significantly improves the efficiency and feasibility of implementing danksharding and Data Availability Sampling (DAS), which enables validators to efficiently validate blob data without downloading entire blocks.

While there is still significant work to be done on the danksharding roadmap, the successful launch of EIP-4844 and blobs shows the potential of Ethereum’s rollup-centric scaling roadmap and a glimpse of what the future of Ethereum could look like.