EIP-4844 Blob Transactions — How Proto-Danksharding Cut L2 Fees by 10x

Before March 13, 2024, using Ethereum Layer 2 networks cost users somewhere between $0.50 and $2 per transaction on average. After March 13, 2024, that cost fell to a fraction of a cent on most L2s. The Dencun upgrade, which implemented EIP-4844, accomplished what years of debate had not: a structural reduction in the cost of L2 data availability. Understanding why requires understanding not just what changed, but what the problem actually was.

## The Calldata Problem

*Rollup* is the term for a class of Layer 2 scaling solutions that execute transactions off Ethereum's main chain but post data about those transactions back to the chain for security. Optimistic rollups (Arbitrum, Optimism/Base) post full transaction data. ZK-rollups (StarkNet, zkSync, Polygon zkEVM) post validity proofs plus transaction data. In both cases, the L2 posts this data to Ethereum mainnet as a transaction *input* — called calldata.

The problem with calldata is that it was designed for a different purpose. Calldata was meant to carry input data for smart contract function calls. It gets stored permanently in Ethereum's transaction history. Rollups repurposed it for data availability, which meant paying for permanent storage when they actually only needed the data to be available for long enough that anyone could verify it and raise a fraud proof (for optimistic rollups) or confirm the validity proof (for ZK-rollups). About 18 days is sufficient.

Permanent storage costs more than temporary availability. Before EIP-4844, rollups were paying for the former when they needed only the latter. This wasn't an efficient use of the fee market.

## What EIP-4844 Actually Does

EIP-4844, authored primarily by Vitalik Buterin, Dankrad Feist, and others, introduced a new transaction type carrying *blobs*. Each blob is a packet of 128 kilobytes of data. Blobs are attached to transactions but are not processed by the EVM — they are not permanent storage. Ethereum nodes store blob data for approximately 18 days, then discard it. The blob data remains accessible during that window for any party that wants to challenge an L2 transaction or verify a proof.

The fee economics are separate from the regular Ethereum gas market. Blobs have their own fee market with their own target (currently 3 blobs per block, maximum 6). When blob demand is low, blob fees are negligible. This separation means that a surge in regular Ethereum usage does not inflate L2 data costs, and vice versa.

The numbers suggest something different from what the critics expected. Early L2 fee data after Dencun showed Base reducing average transaction fees from around $0.30–$0.50 to $0.001–$0.003. Arbitrum showed similar reductions. The 10x headline figure was, in many cases, conservative.

## The Broader Architecture: Why "Proto-Danksharding"?

EIP-4844 is called proto-danksharding because it implements the transaction format and fee structure of full *danksharding* without implementing its most technically demanding component: *data availability sampling* (DAS).

In full danksharding, Ethereum nodes would not need to download entire blobs. Instead, they would use a cryptographic technique where each node samples small random pieces of the blob and can verify, with high probability, that the entire blob is available — without any single node storing the whole thing. This allows the number of blobs per block to scale to 64, 128, or beyond, while each node only does a fraction of the verification work.

DAS requires erasure coding (so that any 50% of blob data is sufficient to reconstruct the whole) and *KZG polynomial commitments*, a cryptographic construction that enables proofs about data content. KZG commitments were introduced into Ethereum by EIP-4844 — they're a component of the blob transaction format. This is why EIP-4844 is proto-danksharding: it establishes the infrastructure that full danksharding will build upon.

It's worth noting that Ethereum's roadmap toward full danksharding involves significant cryptographic and networking engineering. The target of 64 blobs per block — which would multiply L2 throughput again — is not an imminent upgrade. It represents years of additional development.

## What This Means for the L2 Ecosystem

Here's the uncomfortable truth: the fee reduction from EIP-4844 did not benefit all L2s equally or immediately. ZK-rollups, which post more compact validity proofs, were able to take advantage of blob space more efficiently than optimistic rollups posting full transaction data. L2s that had architected themselves around minimal calldata (like StarkNet's compressed transaction representations) benefited differently from those posting raw transaction data.

The reduction in fees also changed the competitive dynamics between L2s and alternative Layer 1s. Before Dencun, the cost argument for using Solana or other alternative chains over Ethereum L2s was partly based on transaction costs. After Dencun, that argument became harder to make for most retail use cases. The fee gap narrowed substantially.

The question of whether further fee reductions will drive meaningfully more demand — more users, more applications, more transaction volume — remains empirically open. Lower fees are necessary but not sufficient for adoption. The user experience of bridging assets between L1 and L2, the fragmented liquidity across chains, and the complexity of the L2 ecosystem for non-technical users remain constraints that fee reduction alone does not solve.

## What Full Danksharding Will Add

Full danksharding's 64-blob target would, at maximum utilization, allow for approximately 1.3 megabytes of L2 data per second on Ethereum — enough to support millions of L2 transactions per day. Combined with improvements in ZK proof generation speed, this begins to approach the throughput required for Ethereum to serve as a settlement layer for internet-scale applications.

The path there is not linear. Peer-to-peer network improvements are required to handle the data routing. Client diversity — multiple Ethereum node implementations — is required to ensure the network isn't dependent on a single codebase handling the new transaction types. Rollup standardization, which would allow shared sequencers and interoperability between L2s, is a separate line of development that EIP-4844 does not directly address.

> **Key Takeaway:** EIP-4844 solved the immediate fee crisis for Ethereum L2s by separating blob data from permanent calldata storage and creating a dedicated fee market. The 10x fee reduction is real and structural. Full danksharding — the end state — requires data availability sampling and will multiply throughput again. The infrastructure decisions made in EIP-4844 were chosen specifically to make that next step possible.

EIP-4844 Blob Transactions — How Proto-Danksharding Cut L2 Fees by 10x

// COMMENTS

ON THIS PAGE