EIP-4844: Shard Blob Transactions
Executive summary
EIP-4844 is the most operationally important Ethereum EIP since the Merge. By introducing a new transaction type carrying ephemeral data 'blobs' that are committed to the chain via KZG polynomial commitments and pruned after roughly 18 days, the proposal severs L2 data availability costs from L1 calldata pricing and delivers what its authors marketed as 'proto-danksharding' — a stepping stone toward full danksharding without the consensus-layer changes that full sharding requires. Activated as part of the Dencun hard fork in March 2024, EIP-4844 reduced L2 transaction costs by 10-100x essentially overnight, reshaped the rollup competitive landscape, and validated the rollup-centric roadmap that had been Ethereum's stated direction since 2020. The proposal is a triumph of incremental engineering: it reuses existing consensus infrastructure, requires only modest client changes, and delivers most of the user-visible benefit of full sharding. DI rates the proposal a ten on impact and an eight on technical elegance.
Background and motivation
By 2021 the rollup-centric roadmap had become Ethereum's official scaling thesis: rather than scaling the L1 directly, push execution to L2 rollups and have the L1 specialise in data availability and settlement. The thesis was sound but bottlenecked by a single quantity: L1 calldata cost. Optimistic and zk-rollups both post their transaction batches to L1 as calldata, paying gas at the same per-byte rate as any other transaction. The result was that L2 transaction fees, while lower than L1, were still dominated by L1 data costs and tracked L1 gas prices closely. During congestion events in 2021, L2 transaction fees on Arbitrum and Optimism could exceed five dollars — undermining the whole point of L2. The full sharding roadmap was always the answer, but full sharding required substantial consensus-layer changes (data-availability sampling, custody-of-bits, complex committee rotation) that were technically and politically distant. Dankrad Feist proposed in 2022 a halfway design — danksharding — that consolidated all data into a single sharded proposer-block rather than the original 64-shard design. Proto-danksharding is the next logical step inward: implement the new transaction format and KZG commitment scheme without yet implementing the data-availability-sampling that full danksharding requires. Validators initially download all blob data, just as they download the rest of the block; DAS is a future upgrade. The political appeal of this staging is that it ships meaningful L2 cost reduction in a tractable timeframe, while keeping the door open to full sharding later. The proposal authors are notably explicit that proto-danksharding is not the destination — it is a foothold.
Core technical contributions
The EIP introduces a new transaction type (0x03) that carries a list of versioned hashes corresponding to KZG commitments to data blobs. Each blob is 4096 field elements of 32 bytes each, totalling 128 KB. The maximum target is six blobs per block (three by default, six maximum), which gives roughly 0.375 MB per block of additional data bandwidth — substantial for L2 rollups but small in absolute terms. Blobs themselves are not stored in the execution-layer state. They are propagated over the consensus-layer p2p network alongside the beacon block, validated by consensus clients, and made available to execution clients for ~18 days (4096 epochs) before being pruned. After pruning, only the KZG commitment remains on-chain; the blob data is the responsibility of any party that wishes to retain it (rollup providers, archival nodes, indexers). The KZG commitment scheme provides two crucial properties. First, it lets validators verify that a blob matches a claimed commitment using a constant-size proof, which is essential for future DAS. Second, it lets execution-layer contracts (e.g. rollup verifier contracts) read the commitment via a new precompile (point_evaluation_precompile) and verify zk or fault proofs against the committed data without having to access the raw blob. The blob fee market is independent of the execution gas fee market: blob_gas_used is tracked separately, with its own EIP-1559-style adjustment mechanism targeting an average of three blobs per block. This decoupling is critical — it means a blob-fee spike does not raise L1 execution gas, and vice versa. The trusted-setup ceremony, completed in early 2023 with over 140,000 contributors, provided the powers-of-tau structured reference string that KZG commitments require, with security holding as long as at least one contributor was honest.
Methodology and rigor
EIP-4844 is the most carefully specified EIP in Ethereum's history, the product of roughly 18 months of design iteration across the Ethereum Foundation research team, client developers, and rollup teams. The specification is broken into four documents: the EIP itself (defining the transaction format and consensus rules), the polynomial-commitments specification (defining KZG operations), the fork-choice updates (defining how blobs interact with attestations), and the consensus-layer specification (defining beacon-block changes). All four were maintained as a single coherent design through multiple devnets (Devnet 4, 5, 6, 7), public testnets (Goerli, Sepolia, Holesky), and shadow forks. The KZG ceremony was itself a substantial engineering project, with formal cryptographic analysis published by the EF cryptography team and external auditors verifying the implementation. From a research-rigor perspective, the proposal benefits from substantial prior work: KZG commitments were published in 2010 (Kate-Zaverucha-Goldberg), polynomial commitment schemes have a deep cryptographic literature, and the data-availability-sampling primitives were extensively analysed in the Al-Bassam et al 'Fraud and Data Availability Proofs' paper (2018). The proposal is essentially an engineering integration of well-understood cryptographic primitives, with the novelty residing in the integration choices (blob size, blob count, fee market parameters) rather than the underlying cryptography. Where the proposal is less rigorous is in its quantitative analysis of post-deployment behaviour. The fee-market parameters were chosen via simulation but were always understood to require post-launch tuning, and the cap of six blobs per block was set conservatively. Both choices have proven approximately correct in practice but were guesses at the time.
Strengths
The proposal's headline strength is operational impact relative to consensus complexity. EIP-4844 ships in a single hard fork (Dencun, March 2024), reuses the existing Beacon Chain p2p infrastructure, and requires no new consensus-layer abstractions beyond blob-sidecar handling. Yet it delivers a 10-100x reduction in L2 fees — the single largest UX improvement in Ethereum's history. Within weeks of activation, Arbitrum and Optimism transaction fees fell below five cents for typical operations, and L2 transaction throughput increased materially. The forward-compatibility is a major design strength: the transaction format, commitment scheme, and pruning rules are all the same shape that full danksharding will use. Implementing data-availability sampling later is a consensus-layer change that does not break user-facing semantics. The independent fee market is another strong design choice. Decoupling blob gas from execution gas means that L2 data costs and L1 execution costs no longer compete for the same finite resource, ending a major source of L2 fee volatility. The KZG ceremony, despite some skepticism about its size and coordination cost, succeeded with extraordinary participation and produced the most-contributed-to trusted setup in cryptographic history — a meaningful security argument by sheer numbers. Finally, the proposal's documentation, public discussion, and devnet rollout were exemplary — it set a new bar for how Ethereum core EIPs are specified and shipped.
Weaknesses and limitations
EIP-4844 is a stepping stone, not a destination, and several of its limitations are explicit design constraints that full danksharding is intended to relax. The most consequential is the 0.375 MB-per-block data ceiling. This is generous compared to pre-Dencun calldata bandwidth but tight compared to long-run L2 demand. By Q1 2026, blob fee spikes during high-activity periods (typically tied to memecoin or NFT bursts on Base or Arbitrum) have become a recurring source of L2 fee volatility, with blob-gas prices occasionally rising 100x intraday. Full danksharding, when shipped, will increase the data ceiling roughly 5-10x but it has not yet shipped — the timeline has slipped from late-2024 estimates to plausibly 2026-27. Second, the trusted-setup requirement is a structural compromise. KZG commitments depend on a powers-of-tau setup, and while the ceremony's design ensured security holding under any-honest-contributor, the dependence on a one-time setup is at odds with the broader ethos of trust-minimisation. Future commitment schemes (FRI-based, transparent setups) would be preferable but were judged not yet practical at proposal time. Third, the 18-day blob retention window has emerged as a practical pain point for rollup operators and indexers who need to reconstruct historical state. Several rollups have built their own archival blob storage (EigenDA, Avail, NearDA-ETH bridges), which arguably re-introduces some of the trust assumptions the proposal aimed to eliminate. Fourth, the proposal's fee-market parameters were tuned for an expected demand profile that has been exceeded. The default-three / max-six blobs-per-block target has been at-or-near maximum during much of 2025, and the EIP-1559-style blob-fee adjustment has lagged demand spikes. EIP-7691 (raising blob count to six default / nine maximum) shipped in Pectra (May 2025) to address this, but additional increases will be needed. Fifth, the precompile gas costs and KZG verification costs are non-trivial and have themselves become a contention point in L2 contract gas budgets.
Subsequent influence
EIP-4844's influence is two-fold: direct technical descendants and indirect ecosystem reshaping. On the technical side, the proposal seeded a family of follow-on EIPs (4844 derivatives 7691, 7762, 7892) addressing parameter tuning, blob data archival, and the path to full danksharding. The KZG point-evaluation precompile has been used by zk-rollup verifier contracts as a primitive for committing to large data without storing it, an application not anticipated by the original proposal. The blob transaction format has been adopted essentially verbatim as the de-facto standard for L2 data posting; even alt-DA layers (EigenDA, Celestia, Avail) frame their security guarantees relative to the EIP-4844 baseline. On the ecosystem side, EIP-4844 catalysed the rollup-centric roadmap from theoretical commitment to operational reality. Pre-Dencun, L2s captured roughly 12% of Ethereum transaction volume; by Q1 2026, that figure is approximately 78%. The cost differential created by 4844 made non-rollup L1 architectures (sidechains, alt-L1s) substantially less competitive for general-purpose smart-contract activity, and several formerly-independent chains (Polygon PoS, BNB Chain) have shipped or announced rollup migrations. The proposal also reshaped the modular blockchain debate: post-4844 Ethereum effectively occupies the data-availability slot at scale, raising the bar for alternative DA layers to differentiate. Ethereum's own DA roadmap — peerDAS in Fusaka, full danksharding in a future fork — is the direct descendant of the staging discipline that 4844 established. Finally, from a research-influence standpoint, the proposal validated that complex consensus changes can be staged into incrementally-shippable EIPs without requiring big-bang upgrades, a methodology that has since been applied to other long-arc Ethereum upgrades (Verkle, EOF, single-slot finality).
How it holds up in 2026
Two years post-activation, EIP-4844 is the clearest case in Ethereum's history of a research-driven upgrade delivering its promised user benefit. L2 fees are 10-100x lower than pre-4844 levels, L2 transaction share has nearly quintupled, and the blob fee market — while not without volatility — has functioned approximately as designed. The proposal's stepping-stone framing has held: the next steps (peerDAS in Fusaka, full danksharding later) build on rather than replace the 4844 design. Where the proposal has aged less well is in its bandwidth assumptions. The 0.375 MB-per-block ceiling is too tight for sustained 2026 L2 activity, and blob-fee spikes have become a regular occurrence. EIP-7691 has lifted the ceiling but full danksharding remains the necessary medium-term answer. The 18-day retention window has driven a market for archival DA services that arguably re-fragments what the EIP attempted to consolidate. The KZG trusted setup, while successful, remains a philosophical compromise that future Ethereum upgrades may revisit if FRI-based or transparent commitment schemes mature. None of these are blocking issues. EIP-4844 was scoped as a halfway step and has held up to its halfway-step ambitions. DI's holistic assessment: among the best-executed major upgrades any blockchain has shipped, validated by user-visible impact, and a model for how complex protocol changes can be staged into deployable units. A pivotal moment in Ethereum's transition from a single-chain to a multi-layer system.
Strengths
- Single-fork proposal delivering 10-100x L2 fee reduction with manageable consensus complexity
- Forward-compatible with full danksharding — same transaction format and commitment scheme
- Independent blob fee market decouples L2 data costs from L1 execution costs
- Rigorous specification across four documents, multiple devnets, and exemplary public process
- KZG ceremony with 140k+ contributors set new bar for trusted-setup security
- Catalysed rollup-centric roadmap from thesis to operational reality (12% to 78% of activity)
Weaknesses
- 0.375 MB-per-block data ceiling is too tight for sustained 2026 L2 demand
- Trusted-setup dependency is a philosophical compromise relative to trust-minimisation ethos
- 18-day blob retention window has driven external archival DA market that re-fragments the design
- Initial fee-market parameters underestimated demand and required EIP-7691 follow-up
- Full danksharding (which this EIP is a stepping stone toward) has slipped multiple times
- Precompile gas costs are a non-trivial overhead in L2 verifier contracts
Key contributions
- New EIP-2718 transaction type (Type-3) carrying KZG commitments to data blobs separately from calldata
- Blob data carried over the consensus-layer p2p network and pruned after ~18 days, reducing long-term storage burden
- Independent EIP-1559-style fee market for blob gas, decoupling L2 data costs from L1 execution costs
- KZG polynomial commitments enabling future data-availability sampling for full danksharding
- Trusted-setup ceremony (KZG Summoning Ceremony) with 140k+ contributors providing the structured reference string
- Forward-compatible design: same transaction format works for full danksharding once DAS is added at consensus layer
TL;DR
The Dencun-fork EIP that severed L2 data costs from L1 calldata pricing and delivered Ethereum's largest user-visible fee reduction. Two years post-activation, has reshaped the rollup landscape and validated the staging discipline behind the rollup-centric roadmap.
Get DeFi Intel research in your inbox
Weekly long-form coverage of papers, incidents, jurisdictions, chains, tokens and the people building them. Free tier covers headlines; Pro adds the analyst-grade breakdowns.
Sources
- eips.ethereum.org/EIPS/eip-4844
- arxiv.org/abs/1809.09044 — Al-Bassam et al, Fraud and Data Availability Proofs (2018)
External references gathered from the body of this brief. Last reviewed 2026-07-15.