EIP-4844 shipped in March 2024 with one clear promise: Layer 2 rollups would post data to Ethereum cheaply, passing those savings to users. For most of 2024, that held. Blob base fees sat near zero, L2 transaction costs dropped to fractions of a cent on chains like Base and Arbitrum, and the narrative around Ethereum as a settlement layer felt vindicated.
Then blob demand started climbing.
By late 2025, the number of blobs per block was regularly hitting the target of three and occasionally pushing toward the six-blob ceiling introduced under the original EIP-4844 parameters. When that happens, the blob fee market - modeled on EIP-1559’s base fee mechanism - starts repricing exponentially. Rollups that had structured their economics around near-zero data costs suddenly faced unpredictable cost spikes. Some, including a few smaller app-specific rollups, responded by batching transactions more aggressively, which introduced latency users noticed.
This wasn’t a failure of the design. It was the design doing exactly what it was supposed to do when capacity is constrained. But it exposed an assumption baked into a lot of L2 roadmaps: that blob capacity would scale faster than blob demand. It hasn’t.
The response is Pectra, which shipped in May 2025 and increased the blob target to six and the max to nine. That gave immediate relief. But the same dynamic is now visible again at higher throughput levels - blob fees are not at crisis levels as of mid-2026, but they’re no longer consistently negligible either.
The deeper issue is that Ethereum’s data roadmap was always going to be a multi-year build. Danksharding - full data availability sampling with potentially hundreds of blobs per block - remains a significant engineering lift. PeerDAS, the intermediate step that enables nodes to verify blob availability without downloading every blob, is progressing through research and testing but doesn’t have a firm mainnet date attached to it.
What this means in practice is that the window between “blobs are cheap” and “blobs are expensive” is narrower than expected, and rollups are building products inside that uncertainty. Base has been explicit that its fee projections depend on continued blob capacity expansion. If PeerDAS slips by another six months, that math changes.
None of this is a reason to write off Ethereum’s L2 scaling thesis. But the blob fee market is already revealing the ordering problem: ecosystem growth is arriving faster than the infrastructure meant to support it, and that gap has a cost.