Rollup Data Availability and Blob Fees

A rollup processes thousands of transactions off the main Ethereum chain and posts only a compressed summary back to it, which is what makes it cheap and fast. But if that summary is all that's published, nobody could reconstruct or challenge the rollup's internal state — so rollups must also publish the underlying transaction data itself, not just the result. This requirement is called data availability: the raw data has to be accessible to anyone who wants to verify it, even if it's never read again.

Before 2024, that data competed for the same expensive block space as regular transactions. EIP-4844 introduced blobs: a separate, temporary data lane attached to each block, priced by its own supply-and-demand market, cheaper because blobs are deleted after about two weeks (verification only needs a short availability window, not permanent storage).

Worked example. Posting 100 KB of rollup data as ordinary calldata might cost the equivalent of several dollars in gas; posting the same data as a blob, priced separately and pruned after ~18 days, can cost a small fraction of that, which is why rollup transaction fees fell sharply after blobs launched.

The pruning window matters because data availability only needs to hold long enough for anyone who wants to challenge or reconstruct rollup state to grab a copy — after that, the data can safely be discarded from the base chain, since a long-term archive isn't required for the chain's own security guarantees. Some rollup designs go further and rely on external data-availability networks rather than blobs at all, trading some of Ethereum's security guarantees for even lower publishing costs.