What is blob transaction?

An Ethereum transaction type introduced by EIP-4844 that carries a large chunk of data priced in its own fee market and deleted by nodes after a fixed retention window.

Blobs exist so rollups can publish their data cheaply without permanently bloating the chain, since the network only needs the data long enough for anyone to check it. Rollup fees fell sharply once blobs replaced calldata.

You never send one. Sequencers do, and the saving reaches you as a lower fee when you transact on a layer 2 network.

How it works

Before EIP-4844, a rollup posted its compressed transaction data as ordinary calldata, which competes with every other transaction for the same block gas and is stored by nodes forever. That was the dominant cost of using a rollup.

A blob is a separate data field attached to a transaction. EIP-4844 specifies each blob as 4,096 field elements of 32 bytes, which is 131,072 bytes, and sets a target of 3 blobs and a maximum of 6 per block in its original form (source: EIP-4844). Later network upgrades have raised those parameters.

Blob gas is its own resource. EIP-4844 states that it is independent of normal gas and follows its own targeting rule similar to EIP-1559, with a separate maximum fee per blob gas field and its own base fee derived from an excess blob gas header value (source: EIP-4844). When blob demand is low the blob base fee collapses to almost nothing, which is why rollup costs became so cheap so quickly.

Retention is deliberately temporary. Consensus nodes keep blob sidecars for a minimum number of epochs, which EIP-4844 describes as around 18 days, after which they may be discarded (source: EIP-4844). The execution layer never sees the blob contents, only a commitment to them.

That retention window is enough for anyone to download the data and check the rollup's state or build a fraud proof, which is the security requirement. Longer-term availability is left to rollup operators and third parties.

Example

Illustrative: a rollup batches 500 user transactions into one blob. If posting that blob costs $6 in total, the data cost per user transaction is $0.012. Under the old calldata approach the same batch might have cost $200, or $0.40 per transaction. The user-visible layer 2 fee falls accordingly, which is the change most people noticed without knowing why. Figures are illustrative; actual costs move with the blob fee market.

Why it matters when you buy

If you use a layer 2 network to move funds cheaply, blob pricing is most of what determines your fee. It is also volatile: when many rollups compete for the same blob space, the blob base fee rises and layer 2 fees rise with it. Check what a withdrawal costs on each network before choosing a route, and see the guide comparing layer 1 and layer 2 and the fee comparison.

Related terms

• data availability — the property blobs are designed to provide
• rollup — the systems that post blobs
• calldata — the expensive alternative blobs replaced
• sequencer — the party that batches and posts
• layer 2 — where the fee saving shows up
• base fee — the same pricing idea applied to ordinary gas

Questions

Can I send a blob transaction myself?

Technically the transaction type is open to anyone, but wallets do not expose it and there is no reason for an ordinary user to send one. It is infrastructure for rollups.

Why did layer 2 fees go up again after falling?

Because blob space is a market. When the number of blobs demanded exceeds the per-block target, the blob base fee climbs, and rollups pass that through.

Is the data lost after 18 days?

The consensus layer stops guaranteeing it. Rollup operators, indexers, and archive services keep copies, so the data usually remains available, but that availability is a service rather than a protocol guarantee.