Rollups spend most of their cost publishing data to Ethereum rather than executing transactions. Blobs exist because that data has requirements ordinary transaction storage does not match.
Rollups must publish data to be verifiable
A rollup executes transactions elsewhere and posts the results to Ethereum. For anyone to challenge or reconstruct that work, the underlying data has to be publicly available.
Originally this data was written as ordinary transaction input, which meant it entered the permanent record and was priced against every other use of block space.
That made rollup costs rise whenever the base chain was busy for entirely unrelated reasons.
The data only needs to be available briefly
What matters is that anyone who wants to verify the rollup can obtain the data during the window in which a challenge is possible.
After that period, the data has served its purpose. Keeping it in every node's permanent storage forever imposes a cost with no corresponding benefit.
Blobs formalise that distinction by carrying data that nodes retain for a defined period and then discard.
A separate market prevents competition for the same space
Blobs have their own capacity target and their own fee that adjusts to blob demand alone.
Ordinary transaction activity therefore does not push blob prices up, and heavy rollup posting does not raise fees for everyone else.
Two independent markets mean each price reflects demand for the resource it actually measures.
Execution never touches the contents
Blob data is not accessible to the virtual machine. Contracts can verify a commitment to a blob but cannot read what it contains.
This restriction is what allows the data to be discarded later, since no contract can depend on it remaining available.
Verification uses a cryptographic commitment posted alongside, so a contract can confirm that particular data was published without holding it.
What this does to rollup pricing
A rollup batches many transactions into one blob, so users share the cost of posting rather than each paying for their own footprint.
When blob demand is below target the price falls towards a floor, which is why rollup fees can be very small in quiet periods and rise when several rollups compete for the same lane.
The number of blobs a block can carry is deliberately conservative, since every node must download and hold them for the retention window even though the data is eventually discarded.
Raising that capacity is therefore the same bandwidth argument as raising block size, applied to a resource whose cost expires rather than accumulating forever.