EIP-4844 shipped in March 2024 with a specific promise: make it cheap for Layer 2s to post data to Ethereum by giving them a dedicated fee market. For several months, it worked. Blob base fees sat near zero, and rollup transaction costs dropped by 80–90% across networks like Arbitrum, Base, and Optimism.
That period is over.
Blob base fees have been spiking intermittently through mid-2026 as aggregate rollup activity has grown faster than the blob target can absorb. The current protocol allows a maximum of six blobs per block, with a target of three. When demand consistently exceeds the target, the fee adjustment mechanism - modeled on EIP-1559 - drives base fees up exponentially. The same logic that made Ethereum’s gas market more predictable is now compressing the window of cheap L2 settlement.
The Target Was Always a Placeholder
The Ethereum development community knew this was coming. The blob target was explicitly designed as a conservative starting point, with the expectation that Danksharding - full data availability sampling across the network - would eventually expand capacity by orders of magnitude. PeerDAS, the intermediate step toward that, is progressing through testing but has not yet shipped to mainnet as of early July 2026.
What wasn’t fully priced in was the speed of L2 adoption. Base alone is regularly among the highest-volume EVM chains by transaction count. Combined blob demand from Base, Arbitrum One, OP Mainnet, Scroll, and newer entrants has at times pushed blocks to their blob ceiling during peak hours.
The result is that L2 sequencers are competing for blob space in a way that wasn’t anticipated at 4844’s launch, and settlement costs - while still cheaper than pre-4844 calldata - are no longer the near-zero figure that made the original upgrade look transformative.
What Pectra Changed (and Didn’t)
The Pectra upgrade, which shipped in May 2025, included EIP-7691, raising the max blob count per block from six to nine and the target from three to six. That provided temporary relief. Fees compressed again through late 2025. By Q2 2026, the same ceiling dynamics are reappearing at the new limits.
This isn’t an indictment of the blob architecture - it’s evidence that Ethereum’s rollup-centric roadmap is generating real demand faster than base-layer capacity is expanding. The problem is that PeerDAS and full Danksharding are multi-year efforts, and in the meantime, L2s are building user bases and fee dependencies on infrastructure that may not hold its cost profile.
Rollup teams are not sitting still. Several are experimenting with data compression improvements and alternative DA layers like Celestia and EigenDA as partial hedges. But the Ethereum-native settlement guarantee that makes rollups credible is precisely what’s becoming expensive again. The tradeoff is getting harder to paper over.