Ethereum Proto-Danksharding: How EIP-4844 Scales Layer 2s

Proto-danksharding (EIP-4844) is live on Ethereum mainnet as of March 13, 2024. It introduced blob transactions, a new data type that lets Layer 2 rollups post data at a fraction of the previous cost, reducing L2 data fees by approximately 90% on day one. Blob capacity has since been expanded in successive upgrades.

Having a clear idea about proto-danksharding means understanding why Ethereum's scaling strategy works the way it does and where it is heading next.

What Is Ethereum Proto-Danksharding?

The name comes from the two researchers who proposed the idea – Protolambda and Dankrad Feist – and was formalized as EIP-4844 (Ethereum Improvement Proposal 4844). The terms proto-danksharding and EIP-4844 refer to the same upgrade.

It went live with the Dencun hard fork on March 13, 2024, becoming one of Ethereum's most significant upgrades since The Merge.

A few things that proto-danksharding is not:

• It is not traditional sharding (splitting the blockchain into parallel chains).
• It does not change how transactions are executed on Layer 1.
• It does not directly lower gas fees for L1 users.

What it does is give rollups a dedicated, cheaper lane to submit their data to Ethereum, and that changes the economics of every Layer 2 built on top of it.

The Problem It Solves: Why Rollup Data Was So Expensive

Here is why that was a problem:

Calldata is permanent. Every byte posted by a rollup was stored forever on-chain, processed by every Ethereum node, and priced accordingly. But rollups only need their data available for a limited window – long enough for fraud proofs or validity proofs to settle. After that, the data serves no purpose on L1.

Rollups competed with L1 activity for the same block space. When Ethereum was congested, like during an NFT mint, an airdrop, or a memecoin surge, L1 gas fees spiked, and L2 data posting costs rose with them. Users on Arbitrum or Optimism felt the impact even though their transactions never directly touched L1.

The structural mismatch was clear. Rollups were paying for permanent storage when they only needed temporary availability.

How Proto-Danksharding Works

Blob-Carrying Transactions (Type 3)

The core innovation of EIP-4844 is a new transaction type, Type 3, that carries one or more blobs.

A blob (Binary Large Object) is a data packet of approximately 128 KB, stored on Ethereum's consensus layer (the Beacon Chain), not the execution layer (the EVM). This distinction is important:

• Blobs are not accessible to the EVM. Smart contracts cannot read blob contents directly.
• Blobs are temporary. They are retained for exactly 4,096 epochs, roughly 18 days, and then pruned from the consensus nodes.
• Each block can carry up to 6 blobs, with a target of 3 per block (as of the original EIP-4844 parameters).

Think of a blob as a detachable trailer on a delivery truck. The truck (the transaction) delivers proof that the trailer existed. After 18 days, the trailer is gone, but the receipt stays on-chain forever.

Because blobs are not stored permanently and do not run through the EVM, they are significantly cheaper than calldata. Rollups use this space to dump compressed batches of their off-chain transactions, then move on.

What stays on-chain permanently is a cryptographic commitment. A short fingerprint proving the blob existed and contained what it claimed. That is where KZG comes in.

My own perspective:

What makes blob-carrying transactions genuinely interesting is the design constraint they work within. Blobs a deliberately weaker data type. They cannot be read by the EVM, they disappear after 18 days, and their availability is guaranteed through cryptographic commitments rather than full replication. That asymmetry is the point. Ethereum is essentially saying: rollups do not need L1 to understand their data; they just need L1 to witness it. Proto-danksharding operationalizes that distinction in a way that preserves decentralization while still giving rollups the cheap, verifiable availability window they need to function securely.

KZG Commitments

Each blob is paired with a KZG commitment – a short cryptographic proof derived from the blob's data. KZG stands for Kate-Zaverucha-Goldberg, after the three researchers who developed the scheme. In practical terms:

• The blob's data is encoded as a polynomial.
• KZG is used to compute a compact commitment to that polynomial.
• Validators can verify the blob's integrity using this commitment without downloading the full 128 KB.

This makes verification efficient at scale, especially important as blob counts increase in future upgrades.

Separate Blob Fee Market

Proto-danksharding introduces a dedicated fee market for blob gas, separate from the regular EIP-1559 execution fee market.

Key properties:

The separation is critical. Before EIP-4844, a DeFi surge on L1 would push up the cost of rollup data posting. With a separate blob fee market, those two costs move independently.

During low-demand periods, which were most of 2024, blob fees sat near the minimum (1 wei per blob gas). During high-demand events like airdrop seasons, blob fees rise on their own curve without dragging up execution gas, and vice versa.

Real-World Impact After Dencun (March 2024)

On March 13, 2024, the day Dencun activated, Layer 2 data posting costs dropped by approximately 90% as rollups switched from calldata to blobs.

Some specific outcomes:

• Base recorded a 224% increase in transaction volume in the weeks following Dencun, driven by lower fees making previously unviable use cases economical.
• Simple token transfers on major rollups fell below $0.01 during normal conditions.
• Arbitrum and Base, which together accounted for roughly 70% of L2 daily transactions in mid-2024, captured the largest absolute fee reductions.
• Small cross-chain transfers (the $50–$500 range) became economically viable for solvers and protocols that previously couldn't absorb destination gas costs.

Blob capacity has also expanded since the initial launch:

• Pectra upgrade (2025): Included EIP-7691, raising the blob target from 3 to 6 per block and the maximum from 6 to 9.
• Fusaka upgrade (late 2025): Introduced PeerDAS (Peer Data Availability Sampling) and a new mini-upgrade format called BPO (Blob Parameter Only) forks, allowing blob capacity to be raised without a full hard fork.
  • BPO1 (December 9, 2025): Target raised to 10, maximum to 15.
  • BPO2 (January 7, 2026): Target raised to 14, maximum to 21.

These two BPO forks alone tripled blob capacity within a single month.

What did not change: L1 calldata fees remained unchanged. Users transacting directly on Ethereum mainnet saw no fee reduction. The benefits of EIP-4844 flow exclusively through Layer 2 rollups.

Proto-Danksharding vs. Full Danksharding

Proto-danksharding and full danksharding share the same architectural vision but differ significantly in what they actually deploy.

The key missing piece in proto-danksharding is Data Availability Sampling (DAS). Today, consensus nodes still download every blob in full. Full danksharding replaces this with a probabilistic sampling system: each node downloads only small random pieces of blob data, and the statistical guarantee ensures that if enough samples are available, the full data exists.

• DAS is what allows blob capacity to scale to 64–128 per block without requiring nodes to store proportionally more data, preserving decentralization at a much higher throughput ceiling.
• PeerDAS, introduced in Fusaka, is an intermediate implementation of this concept that production validators can run today.

Why Proto-Danksharding Matters for Ethereum's Future

Ethereum's rollup-centric roadmap treats Layer 2s as the primary execution environment, while L1 serves as the settlement and data availability layer. For this model to work, L1 data must be cheap enough that rollups can operate economically. EIP-4844 was the proof-of-concept that this architecture actually functions.

Two years after Dencun, Arbitrum and Optimism process more daily transactions than Ethereum mainnet – a reversal that underscores how central rollups have become.

The ETH deflation trade-off

One consequence of the blob fee market is less ETH burn. Blob transactions use a separate fee market with a much lower base fee than execution gas. As rollup activity grew and more data moved to blobs, total ETH burned through EIP-1559 declined. This has shifted the conversation around ETH's monetary policy: less "ultrasound money," more "digital infrastructure with protocol yield."

The data availability competition

EIP-4844 also changed the competitive landscape for data availability layers. Projects like Celestia, EigenDA, and Avail positioned themselves as alternatives to Ethereum for rollup data posting. Proto-danksharding and the subsequent blob capacity expansions made Ethereum a more competitive option for rollups that want to keep their data layer and security layer unified.

Sources and Further Reading

• Ethereum.org – "Danksharding" https://ethereum.org/roadmap/danksharding/
• eip4844.com – "EIP-4844: Proto-Danksharding" https://www.eip4844.com/
• Ethereum EIPs GitHub – "EIP-4844: Shard Blob Transactions" https://eips.ethereum.org/EIPS/eip-4844
• ConsenSys Teku – "Proto-Danksharding" https://docs.teku.consensys.io/concepts/proto-danksharding
• Vitalik Buterin – "Proto-Danksharding FAQ" https://notes.ethereum.org/@vbuterin/proto_danksharding_faq