EIP-4844, known as proto-danksharding, is the 2024 Ethereum upgrade that gave Layer-2 networks a cheap new place to store their data called blobs, which cut the fees people pay on rollups like Arbitrum and Base by roughly 90 percent. It shipped in the Dencun upgrade on March 13, 2024, and this guide explains how it works.
What is EIP-4844, also called proto-danksharding?
EIP-4844 is the Ethereum change that introduced "blobs," a low-cost lane for the data that Layer-2 networks post back to Ethereum, and it is the single biggest reason rollup fees fell from dollars to cents. It went live in the Dencun upgrade on March 13, 2024 (source: Dencun mainnet announcement, ethereum.org). The nickname "proto-danksharding" signals that it is an early, partial version of a larger scaling plan called danksharding.
To see why it mattered, you need the problem it solved. Ethereum is slow and expensive to use directly, so most everyday activity has moved to Layer-2 networks, also called rollups. A rollup bundles thousands of transactions, processes them cheaply on its own, then posts a compressed record back to Ethereum for safekeeping. The catch was the cost of posting that record. Before 2024, rollups had to store their data in a part of every block that stays on-chain forever, and paying for permanent storage made each posting expensive. Those costs were passed straight to users.
EIP-4844 fixed this by adding a second, cheaper place to put that data. Instead of paying for permanent block space, rollups can now attach their data as a blob, a large packet that Ethereum keeps for only a short time and then deletes. That one change is the heart of the upgrade. The rest of this guide explains what a blob actually is, why it made fees collapse, and where the technology is heading next.
What is a blob, and how is it different from calldata?
A blob is a large packet of data, about 128 kilobytes in size, that a Layer-2 attaches to an Ethereum block but that the network keeps for only about 18 days before deleting it. The old method, called calldata, stored the same data permanently inside the block, where every Ethereum node had to download and keep it forever. Blobs are cheaper because Ethereum no longer has to store them for good.
The difference comes down to how long the data has to live. A rollup only needs its data to be available for a short window, long enough for anyone to check its work and challenge it if something looks wrong. After that window, the data has served its purpose. Permanent storage, the way calldata works, was paying for decades of safekeeping that nobody needed. Blobs match the price to the actual need, which is short-term availability (source: EIP-4844 specification).
Feature | Calldata (old method) | Blob (after EIP-4844) |
|---|---|---|
How long it is kept | Forever, in every block | About 18 days, then deleted |
Who stores it long-term | Every Ethereum node, permanently | Nobody, after the window passes |
What it costs | Competes with all regular gas | Has its own separate, cheaper fee |
Size | Small, expensive per byte | About 128 KB, cheap per byte |
The path a blob takes is short and simple. A rollup gathers thousands of user transactions off-chain. It attaches a compressed record of them as a blob to an Ethereum block. Ethereum nodes hold that blob for about 18 days, long enough for anyone to verify the rollup's work, then prune it.
The blob still gets checked and secured by Ethereum while it is live, so the safety of the data does not change. What changes is the bill. Rollups stopped renting permanent space they did not need and started renting a short-term locker instead. The Consensys breakdown of the upgrade frames blobs as exactly this: a purpose-built, temporary data store for rollups (source: Dencun upgrade and EIP-4844, Consensys). This is also why people compare the blob change to Ethereum's broader fee design, which our explainer on why Bitcoin fees get high shows is always about competing for limited block space.
Why did Layer-2 fees drop so much after EIP-4844?
Layer-2 fees dropped because blobs gave rollups their own separate, low-cost fee market instead of forcing them to bid against every other transaction for permanent block space. With supply of cheap data space suddenly much larger, the price to post a rollup batch collapsed, and in the weeks after Dencun major Layer-2 fees fell by roughly 90 percent or more.
Here is the mechanism in plain terms. Before the upgrade, a rollup posting its data competed in the same auction as someone swapping tokens or minting an NFT. When the network was busy, that auction got expensive, and rollups had to pay up. Blobs created a dedicated lane with its own pricing, so a rollup posting data no longer competes with ordinary transactions. The blob lane also offered far more room, which kept the per-blob price low.
A worked example shows the scale. Suppose a rollup batch that records a few thousand user transactions cost the rollup around 100 dollars to post using the old calldata method. Spread across those users, each person paid a small share, but on a busy day that share could still be 1 to 2 dollars per swap. After EIP-4844, the same batch posted as a blob might cost the rollup a few cents to a couple of dollars total, so each user's share drops to a fraction of a cent. That is why a swap that once cost over a dollar on an L2 now often costs a few cents, a sharp drop in the data-availability fees rollups pay (source: Impact of EIP-4844 on Ethereum, Hacken).
The before-and-after picture is easiest to read in a table.
Cost | Before EIP-4844 (calldata) | After EIP-4844 (blobs) |
|---|---|---|
Rollup cost to post one batch | Often 100 dollars or more on busy days | A few cents to a couple of dollars |
A user's share per L2 swap | Often 1 to 2 dollars | A fraction of a cent to a few cents |
From Blofin's operational perspective, the cost a user feels when bridging funds or trading on a Layer-2 traces straight back to blob space. When blobs are plentiful and cheap, those withdrawals and swaps clear for pennies, and when blob demand spikes, we see the same L2 costs tick up. The price of using a rollup is, in large part, the price of blob space passed along to you. For the wider picture of what crypto fees include, our guide to crypto fees for investors breaks down the costs that stack up around a trade.
Did EIP-4844 make Ethereum mainnet cheaper too?
No, not directly. EIP-4844 lowered the cost of Layer-2 transactions, not the cost of using Ethereum's main layer itself. Blobs are a separate lane built specifically for rollup data, and they do not change the regular gas price you pay to send ETH or use an app directly on Ethereum mainnet. The relief flows to users through cheaper rollups, not through cheaper base-layer gas.
This is a common point of confusion worth clearing up. If you send a transaction directly on Ethereum mainnet, your fee still depends on regular gas demand, the same as before 2024. What EIP-4844 did was make the cheap path, using a Layer-2, dramatically cheaper. The intended design of Ethereum is that most people transact on rollups and only settle to mainnet when needed, so making rollups cheap is the point. Mainnet gas does still drift lower during quiet periods, but that is ordinary supply and demand, not a result of blobs. How base-layer gas pricing works is a separate topic, covered in our standalone guide to Ethereum gas, and the contrast with Bitcoin's fee approach appears in Bitcoin versus Ethereum.
What is the difference between proto-danksharding and full danksharding?
Proto-danksharding, which is EIP-4844, shipped the blob format and its cheap fee market, but it did not yet add the technique that lets Ethereum scale to a huge number of blobs. Full danksharding adds that missing piece, called data availability sampling, which will let the network safely carry many times more blobs per block. Proto-danksharding is the foundation; full danksharding is the finished building.
The dividing line is how nodes confirm the data is really there. Today, with proto-danksharding, every node still downloads every blob to be sure the data is available. That works, but it limits how many blobs a block can safely hold, because asking every node to download more and more data would shut out anyone without a powerful computer. Data availability sampling solves this. Instead of downloading a whole blob, each node checks many small random pieces of it. If enough random samples are present, the full data is almost certainly available, and no single node has to carry the whole load (source: Danksharding, ethereum.org).
That shift is what makes real scale possible. Once nodes can verify data by sampling, Ethereum can raise the blob count far beyond today's levels, with a long-term target of 64 or more blobs per block under full danksharding (source: Data availability, ethereum.org). More blob space means even lower Layer-2 fees, even when the network is busy. The work is being delivered in stages rather than one giant release, which is why you hear about steady upgrades rather than a single "danksharding day."
What has changed for blobs since Dencun?
Blob capacity has grown in steps since 2024, and each step adds more cheap room for Layer-2s. Dencun started with a target of 3 blobs per block and a maximum of 6. The Pectra upgrade in May 2025 raised that to a target of 6 and a maximum of 9. The Fusaka upgrade in December 2025 then introduced data availability sampling and pushed the numbers higher still.
The reason for raising the limits is simple: Layer-2 use grew so fast that blocks kept hitting the old blob ceiling, which pushed blob fees back up. Increasing the target and maximum gives rollups more room and keeps the cheap lane cheap. Fusaka was the key step, because it shipped a first version of data availability sampling called PeerDAS, which lets nodes verify blobs by sampling rather than downloading every one (source: EIP-7594 PeerDAS specification). That made it safe to raise blob counts without forcing every node to download more data.
Upgrade | Date | Blob target / maximum per block |
|---|---|---|
Dencun (EIP-4844) | March 2024 | 3 / 6 |
Pectra | May 2025 | 6 / 9 |
Fusaka, then staged increases | December 2025 onward | 10 / 15, rising to 14 / 21 |
After Fusaka, Ethereum began using small "blob parameter only" upgrades to step the numbers up safely, reaching a target of 10 and maximum of 15, then a target of 14 and maximum of 21 in early 2026 (source: Fusaka mainnet announcement, ethereum.org). These figures keep moving as the network scales, so treat any specific number as a snapshot. The direction is steady: more blob space over time, with full danksharding as the destination. Where each of these upgrades sits in the bigger plan is covered in our overview of the Ethereum roadmap.
What does proto-danksharding mean for a regular user?
For everyday users, proto-danksharding is the reason transacting on a Layer-2 is cheap, and you benefit from it automatically without doing anything. You never create or pay for a blob yourself. Your rollup does that behind the scenes, and the savings show up as low fees when you swap, bridge, or send on networks like Arbitrum, Base, or Optimism. There is no setting to change and no action to take.
The practical takeaway is that the cheap way to use Ethereum is through a rollup, and blobs are what make that path affordable. If a transaction directly on Ethereum mainnet looks expensive, moving the same activity to a Layer-2 is usually far cheaper, and that gap exists because of EIP-4844. It is also worth knowing that blob fees can rise during very busy periods, since the blob lane has its own supply and demand, so L2 fees are low but not perfectly fixed. As Ethereum adds more blob space through danksharding, the expectation is that the cheap path stays cheap even as more people use it. For deciding where you actually transact, our comparison of centralized and decentralized exchanges is a useful next step, since many DEX trades now happen on these low-fee Layer-2s.
Frequently asked questions
Are blobs stored on Ethereum forever?
No. Blobs are deliberately temporary. Ethereum keeps each blob for about 18 days, then deletes it, which is exactly why blobs are so much cheaper than the old calldata method that stored data permanently. The short window is long enough for anyone to download the data, check the rollup's work, and raise a challenge if something is wrong. After that, the data is no longer needed on-chain, so keeping it would be wasted cost. Rollups keep their own copies of the underlying data.
Do I pay blob fees directly when I use a Layer-2?
No. You never pay a blob fee yourself. The Layer-2 network you are using pays to post its data as blobs, then spreads that cost across everyone in the batch. You feel it indirectly as part of your low Layer-2 transaction fee. This is why an L2 swap might cost a few cents: your tiny share of the blob cost, plus the rollup's own small fee. When blob space is cheap and plentiful, your share is almost nothing.
Can blob fees spike like regular gas?
Yes. Blobs have their own separate fee market, so when many rollups compete for blob space at once, the blob fee rises and Layer-2 fees tick up with it. This was happening before the Fusaka upgrade, when Layer-2 demand kept hitting the blob ceiling. Raising the blob target and maximum, and adding data availability sampling, is how Ethereum keeps that lane from getting congested. So L2 fees are usually very low, but they are not a fixed price.
Is proto-danksharding the same as sharding?
No, and the difference matters. Older Ethereum plans involved "sharding," splitting the whole network into many parallel chains to spread out the work. Ethereum dropped that approach in favor of letting rollups handle scale, with the base layer focused on storing their data cheaply. Danksharding is named after that history but is really about scaling data availability for rollups, not splitting execution into shards. Proto-danksharding, EIP-4844, is the first step of that data-focused plan.
Which networks actually benefit from blobs?
Rollups, the Layer-2 networks that post their transaction data back to Ethereum, are the direct beneficiaries. That includes well-known networks such as Arbitrum, Optimism, Base, and zkSync. Because their biggest cost was posting data to Ethereum, blobs cut that cost sharply and let them pass the savings to users. Networks that do not settle their data to Ethereum do not use Ethereum blobs, so the benefit is specific to Ethereum rollups.
Is danksharding finished?
No. Proto-danksharding shipped in 2024, and the Fusaka upgrade in December 2025 added a first version of data availability sampling called PeerDAS, but full danksharding is still being built. The plan is to keep raising blob capacity in safe steps toward a long-term target of 64 or more blobs per block. Each step lowers Layer-2 fees further and is delivered as a separate upgrade, so the work continues across several releases rather than arriving all at once.
Researched and written by the Blofin Academy editorial team with AI-assisted drafting. Primary sources include the Ethereum.org Dencun and danksharding documentation, the EIP-4844 specification at eips.ethereum.org, and the Ethereum Foundation's Fusaka mainnet announcement, with fee-impact framing cross-checked against the Consensys EIP-4844 breakdown. All facts independently verified against cited sources current as of June 2026.
This article is for informational purposes only and does not constitute financial advice, investment guidance, or a recommendation to buy, sell, or hold any digital asset. Blob capacity, fee levels, and upgrade dates change frequently; the figures here are snapshots current as of 2026. Cryptocurrency markets involve significant risk and you should conduct your own research and consult qualified professionals before making investment decisions. Blofin Academy content reflects the state of public information at time of publication; protocol parameters, fees, and ecosystem data change frequently.