What Is EIP-4844? Ethereum's Cheap-Fee Scaling Milestone

EIP-4844 introduced blob transactions and cut the cost of putting L2 data on chain by 10–100x — that single sentence explains why fees on Arbitrum, Optimism and Base all collapsed after the March 2024 Dencun upgrade. It did not speed up Ethereum mainnet itself; it gave every rollup a cheap data substrate to sit on.

First, the L2 bottleneck was never execution — it was data

People often assume an Optimistic Rollup or a ZK Rollup is “done” once it batches transactions on its own L2. It is not. After a rollup computes state on L2, it still has to write that transaction data back to Ethereum mainnet, otherwise no one can independently reconstruct state or verify a fraud proof. Rollup economics has a name for this layer: data availability (DA), and before EIP-4844 the whole L2 story was bottlenecked exactly there.

That mainnet writeback used to ride in the normal calldata field of an ordinary transaction. Three problems came with that:

• Permanent storage: every node had to keep it forever.
• Priced in normal gas: whenever mainnet got congested, L2 fees jumped right with it.
• Competing for execution bandwidth: rollup data fought for the same lane as regular transfers.

The result was that the “cheap L2” feeling was always being eroded; during a hot mint, a single swap on an L2 could still cost several dollars.

(/uploads/20260529/1780060829977-2147.png)

What EIP-4844 actually did: a dedicated lane for rollups

EIP-4844, also known as Proto-Danksharding, is the precursor of the larger Danksharding design. Its core change is a single new transaction type: the Blob-carrying Transaction.

A blob is a fixed-size (~128 KB) “data packet” broadcast alongside the transaction, with a few key properties:

• It does not enter the EVM. Contracts cannot see blob bytes, only a short KZG commitment.
• It is temporary. After about 18 days the consensus layer can drop it; no permanent disk burden.
• It has its own fee market. Blobs carry blob_gas priced by a separate blob_base_fee, decoupled from normal gas.

Effectively, rollups got a dedicated express lane for data. Regular transaction congestion stops splashing onto L2 data. From a node operator’s point of view, blob load is handled by the consensus layer rather than execution: the bandwidth of pulling and gossiping blobs is unrelated to the CPU cost of running contracts, so Ethereum nodes can absorb meaningfully more data without sacrificing decentralization.

Blob data vs calldata: where it matters

The key insight: data availability only needs to hold during the challenge window. Optimistic rollups have ~7 days of fraud-proof window; ZK rollups don’t need the raw data at all after verification. So “permanent storage” was always an overpromise, and the temporary nature of blobs matches what rollups actually need.

Real impact on L2 fees

On the day Dencun activated, average fees on Base dropped from the $0.20–$0.50 band straight into single cents, with Arbitrum and Optimism following. This was not a subsidy — rollups simply migrated their data from expensive calldata to cheap blobs.

But “permanently free” was never the deal — blob_base_fee also has EIP-1559-style automatic adjustment:

• Blob usage below target (3 per block): the rate falls exponentially, L2s become extremely cheap.
• Blobs persistently maxed (6 per block): the rate rises exponentially, L2 fees climb back up.

The 2024 Bitcoin Runes and inscriptions craze, and the early-2025 wave of memecoins and L2 airdrops, both spiked blob_base_fee by dozens of times. Blob capacity is still finite, and the real scaling work waits for full Danksharding. For how the L2 majors actually divide up this cheap substrate, see Arbitrum vs Optimism vs Base.

What comes next: from 6 blobs to 64

EIP-4844 is just the “Proto” version. Full Danksharding aims for:

• Blob count from 6 up to 64: roughly 8 MB of data bandwidth per block.
• Data Availability Sampling (DAS): nodes only sample randomly, no need to download every blob.
• PBS (Proposer-Builder Separation): further separating block building from proposing, resisting MEV centralization.

The Pectra upgrade on 7 May 2025 already raised the target/max blob count from 3/6 to 6/9 via EIP-7691 — a meaningful step toward Danksharding. The later Fusaka upgrade will introduce PeerDAS and actually land data availability sampling.

What it means for developers and users

For developers, EIP-4844 did not change the contract interface, but it dramatically lowered the operating cost of L2 deployment. Before blobs, rollup operators were spending tens to hundreds of ETH a day writing data to mainnet — the biggest line in their P&L. Once blobs cut that one to two orders of magnitude, third-party Rollup-as-a-Service platforms (Conduit, Caldera, Gelato) finally had real margin, which spawned a wave of new chains. Projects built on shared frameworks like OP Stack benefited the most, because they already treated DA as a pluggable module.

For ordinary users the change is even more direct:

• On-chain swaps, mints and transfers on L2 are once again “daily-affordable” — a basic action dropped from $1–2 to one or two cents.
• Small payments (<$10) became economically viable; high-frequency apps such as tipping, on-chain mini-games and message-based payments suddenly had room to be commercial.
• Round-trip costs between L2s fell, letting capital move flexibly across chains and lowering the bar for arbitrage and structured strategies.
• Identity and credential apps benefit most: signing in, checking in, claiming an SBT no longer means weighing gas, and developers are willing to build “write to state often” products.

(/uploads/20260529/1780060866935-7027.png)

Blobs are Ethereum’s cheap substrate for L2s

EIP-4844 didn’t make mainnet faster. What it did was move the data rollups truly need from a crowded execution lane into a dedicated data channel. L2s no longer pay for permanent storage, no longer fight transfers for gas, and their fees are decoupled from mainnet congestion.

Next, Danksharding will widen this channel from 6 lanes to 64, and DAS will keep node verification costs flat. Until then, blobs have already pulled L2 UX into a brand-new price tier — which, by itself, is the first cornerstone of Ethereum’s “rollup-centric roadmap.” Compressed into four numbers: each blob ~128 KB, target 6 per block today, auto-pruned after ~18 days, and a fee market independent of gas. Those four parameters define the cost ceiling of L2 right now, and however far Danksharding pushes them, the core idea no longer changes.