What is EIP-4844?

EIP-4844 is the Ethereum Improvement Proposal that introduced blob-carrying transactions. These type 3 transactions carry large pieces of data called blobs alongside the block on the consensus layer, which runs Ethereum's proof-of-stake system. Blob data stays outside the execution layer where smart contracts run.

Proto-Danksharding takes its name from the researchers who proposed it, Protolambda and Dankrad Feist, as ethereum.org explains. "Proto" identifies it as the first stage of full Danksharding, a later design that removes the need for any single node to download all blob data.

The proposal describes rollups as the only trustless way to scale Ethereum in the short and medium term. Its initial blob capacity target was about 0.375 MB per block, with a limit of about 0.75 MB.

By introducing the exact transaction format intended for full Danksharding, EIP-4844 required rollups to make only one major change to how they post data. Subsequent capacity increases did not require another transaction type.

Why Rollups Needed a New Type of Data

A rollup executes transactions on a separate network, then posts a compressed record of them to Ethereum. Anyone can use that record to rebuild and check the rollup's state. This guarantee is called data availability.

For optimistic rollups, the data allows anyone to challenge an incorrect state during the dispute window. ZK rollups prove each state change with a validity proof, but users still need the underlying data to confirm their balances and exit if the operator stops working. Both rollup designs require that data for only a limited time.

Before March 2024, rollups posted these records as calldata, the input field of a normal transaction. Every Ethereum node processes calldata and stores it permanently, so rollups paid for permanent storage when a few weeks of availability would have been enough.

Sharing the same gas market as other Ethereum transactions added another cost. Busy periods on mainnet pushed up rollup data fees, which users then paid. EIP-4844 addressed both problems by giving blobs their own price and a fixed deletion date.

How Blob Transactions Work

A blob transaction resembles a normal Ethereum transaction with additional data attached. The EIP-4844 specification establishes the core rules:

• Fixed size: A blob holds 4,096 field elements of 32 bytes each, totalling 131,072 bytes (128 KiB). Senders pay for the entire blob even when they use only part of it.
• Temporary storage: Consensus nodes retain blobs for 4,096 epochs, about 18 days, before deleting them. Rollup operators, archive services and users can retain their own copies for historical use.
• Hidden from smart contracts: The Ethereum Virtual Machine (EVM), which runs smart contracts, cannot read blob contents. Through the BLOBHASH opcode, contracts can access only a 32-byte versioned hash of each blob.
• Cryptographic commitments: Each blob includes a KZG commitment, a short cryptographic fingerprint of its data. Rollups can use a precompiled contract at address 0x0A to prove that a specific value belongs to a blob, at a cost of 50,000 gas.
• Separate fee that is burned: Blob gas has a separate base fee. The protocol burns the full blob fee and does not refund it if the transaction fails.

KZG commitments rely on secret random numbers that nobody can know. To generate them, Ethereum held a public ceremony with more than 140,000 contributions, the largest of its kind, according to ethereum.org. The setup remains secure as long as at least one participant destroyed their secret value.

Nodes verify each blob once without executing it or storing it permanently. That limited workload allows the network to price blob data far below calldata.

How the Blob Fee Market Sets Prices

Blob pricing follows the same principle as EIP-1559, which sets Ethereum's normal base fee. The protocol specifies a target and a maximum number of blobs per block. Usage above the target raises the blob base fee, while usage below it lowers the fee.

A running total called excess blob gas tracks this difference, and the base fee grows exponentially with that total. At launch, a block containing the maximum of 6 blobs increased the fee by about 12.5%. An empty block reduced it by about 11.1%.

Pectra's EIP-7691 changed the target-to-maximum ratio from 1:2 to 2:3 and adjusted the fee's update speed. A full block now raises the fee by about 8.2%, while a block with no blobs lowers it by about 14.5%.

Low demand exposed a weakness in the original design. Whenever supply exceeded demand, the blob base fee often dropped to its minimum of 1 wei, the smallest unit of ETH. Recovering from that floor after a sudden demand increase took more than an hour of near-full blocks before the fee reached a market price again, the authors of EIP-7918 noted.

Blob Capacity Since Dencun

Ethereum has increased blob capacity three times since March 2024, raising both the target and the maximum number of blobs per block:

The current target is about 4.7 times its launch level. Maximum capacity is 3.5 times higher, allowing a full block to carry about 2.6 MiB of blob data.

Alongside its capacity increase, Pectra introduced EIP-7623. For data-heavy transactions, it set minimum calldata prices of 10 gas per zero byte and 40 gas per non-zero byte. Calldata became a poor choice for rollup data, encouraging remaining users to switch to blobs.

Fusaka's EIP-7892 established Blob Parameter Only (BPO) forks. These change only the blob target and maximum, plus the speed of blob fee adjustments. Client teams implement the changes through configuration rather than new code, allowing capacity increases without waiting for a full named upgrade.

As confirmed in the Ethereum Foundation's mainnet announcement, the first two BPO forks were scheduled together with Fusaka. Both activated on time.

How PeerDAS Changed Blob Verification

Every consensus node downloaded every blob before Fusaka, making home validators' bandwidth the upper limit on capacity.

PeerDAS (EIP-7594) replaces those full downloads with data availability sampling. Each blob is first extended through erasure coding, which adds redundant data so the original can be reconstructed from part of it. The protocol divides the extended data into 128 columns. Individual nodes store and check only some of them.

Under current parameters, a node needs about one eighth of the total data to check availability, according to EIP-7594. Collecting at least half of the columns allows any node to reconstruct the complete data. As the network grows, the chance of an attacker hiding data while passing these checks becomes negligible.

Rollups sending blobs also take on different work under PeerDAS. They must compute 128 cell proofs per blob and attach them to the transaction, moving expensive computation away from block producers. EIP-7594 limits each transaction to 6 blobs.

The Ethereum Foundation says PeerDAS enables an 8x increase in theoretical blob capacity. An earlier protocol update outlined a path from 6 to as many as 48 blobs per block using this architecture.

EIP-7918 and the Blob Fee Floor

Another Fusaka proposal, EIP-7918, ties blob fees to the execution base fee. Its reserve price requires one blob to cost at least as much as 8,192 units of normal execution gas. Each blob consumes 131,072 units of blob gas, making the reserve blob base fee one sixteenth of the execution base fee.

The reserve price aims to restore a working price signal. When execution costs account for most of a rollup's total expense, changes to the blob base fee barely affect demand, and that fee falls to 1 wei. The reserve also charges blob users for part of the computing work needed to verify KZG proofs.

In absolute terms, the floor remains small. An execution base fee of 1 gwei puts the minimum cost of a blob at 8,192 gwei, or about 0.0000082 ETH. High blob demand leaves the rule without effect because the market fee already exceeds the reserve.

Fidelity Digital Assets modelled the reserve price as though it had applied since Dencun. The adjusted fee would have exceeded the observed fee on 93% of days, adding about 24,641 ETH in revenue up to October 28, 2025. Its estimate valued that amount at about $78.6 million at an ETH price of $3,984.

Blob Demand in 2026: Unused Capacity and Network Limits

After the BPO forks, blob supply outpaced demand. Core developers agreed in January 2026 that a third BPO fork was not a priority until usage increased, the Ethereum Foundation reported.

Usage subsequently reached new highs. On September 3, 2026, it hit a record three-day average of 5.9 blobs per block and a daily average of 6.7, Bankless reported. Those figures still represented only 40% to 50% of the 14-blob target. Blocks containing the full 21 blobs remained rare.

Further expansion also depends on network performance. MigaLabs research found that blocks carrying many blobs were more likely to cause the following block to be missed, DL News reported. In early 2026, developers reached rough agreement that blob data needed to spread across the network faster before capacity rose again.

The next proposed increase would set a target of 21 blobs and a maximum of 32. Developers are weighing lower rollup fees against the added bandwidth load, alongside the large gas limit increase planned for Glamsterdam. No third BPO fork has a mainnet date as of late October 2026.

Who Captures the Value: Rollup Fees, Blob Revenue and ETH Supply

EIP-4844 met its central goal of making Layer 2 transactions cheap. Base, Arbitrum and OP Mainnet now post data at a small fraction of their former cost. Our Layer 2 rankings show how many networks depend on that pricing.

Low fees also restrict Ethereum's earnings. Over the year to late 2025, Base paid about $5.2 million in blob fees while earning about $94 million from user transaction fees, according to Fidelity. Most of the value generated by its activity stayed with the rollup.

Because blob fees are burned, this has a direct effect on ETH supply. Execution and blob fee burns removed about 16,525 ETH between January 1 and October 9, 2026, offsetting only 2.07% of gross issuance, TokenPost calculated. Net supply increased by about 778,000 ETH during that period. The current balance between burn and issuance is available on our ETH gas tracker.

Supporters argue that cheap blobs expand total activity. In their view, higher capacity and the EIP-7918 reserve price could make blob fees a larger revenue source over time. Critics point to rollups retaining most fees while paying little to Ethereum, weakening the connection between network use and the price of ETH.

EIP-4844 and Ethereum's Changing Roadmap

The rollup-centric roadmap behind EIP-4844 assigned Ethereum mainnet the role of maintaining security while Layer 2 networks handled most activity. That plan has changed in 2026.

On February 3, 2026, Vitalik Buterin wrote that Layer 2 decentralisation had progressed far more slowly than expected, even as mainnet itself was scaling, The Block reported. He said the original vision for Layer 2 networks and their role in Ethereum "no longer makes sense" and called for a new path.

Both directions appear in the Ethereum Foundation's 2026 protocol priorities. Mainnet execution and blob capacity now sit within a single Scale track, whose Glamsterdam work includes further blob parameter increases. During 2025, the mainnet gas limit doubled from 30 million to 60 million. That limit measures how much computation fits into one block.

Competition also comes from dedicated data availability networks. Celestia and EigenDA, which uses ETH restaked through EigenLayer, offer more raw capacity at lower cost under different security models. Fidelity warned that raising blob prices could drive some rollups toward these cheaper alternatives.

What Comes Next for Blobs

Glamsterdam, the next major upgrade, activated on the Sepolia test network on October 6, 2026, in line with the Ethereum Foundation's schedule. Hoodi test network activation is scheduled for October 26, 2026. A mainnet date has yet to be agreed, Yellow reported.

Several Glamsterdam changes prepare the network to handle more blobs:

• Enshrined proposer-builder separation (EIP-7732): This brings block building into the protocol and extends the time available to spread data across the network from about 2 seconds to about 9 seconds, ethereum.org explains. A new committee checks that each block's blob data arrived on time.
• Sparse blobpool (EIP-8070): Each execution client has a 15% probability of downloading a full new blob. Otherwise, it samples only its assigned columns. The proposal estimates that this reduces average blobpool bandwidth by about 4x.
• Higher calldata floor (EIP-7976): Minimum calldata pricing rises to 64 gas per byte. Stronger incentives to use blobs are among the proposal's stated goals.

Planned to follow Glamsterdam, Hegota has fork-choice enforced inclusion lists (FOCIL) as its headline feature. The Ethereum Foundation includes censorship resistance for blobs among the research areas associated with that work.

The long-term goal remains full Danksharding. Ethereum.org places it several years away and expects it to support more than 100,000 transactions per second across rollups. PeerDAS and BPO forks, along with Glamsterdam's networking changes, are steps toward that design.

Final Thoughts

EIP-4844 gave rollups the cheap, temporary data they needed. Blobs cut their largest cost by more than 90% overnight and preserved the path to full Danksharding.

Most remaining challenges are economic and political. Blob space exceeds rollup demand, leaving blob fees with little contribution to ETH burn. Meanwhile, Ethereum is reviewing the role of rollups in its roadmap.

We expect demand and the network's ability to carry larger blocks to determine the pace of future BPO forks. Until usage reaches the current target, the most important blob metric is how much of the 14-blob target rollups consume each day.