Since EIP-4844 shipped in March 2024, blob fees have settled into a pricing dynamic that nobody fully predicted. Rollups like Arbitrum, Base, and Optimism moved aggressively to blob-based data posting and saw transaction costs drop by 80–90% almost immediately. That part went as designed. What’s less discussed is what happens when blob demand spikes and the separate blob fee market - running in parallel to the legacy execution gas market - starts to price out smaller rollups during congestion windows.
Blob space is capped at six blobs per block under current parameters, with a target of three. When multiple major rollups post simultaneously during high-activity periods, blob base fees can spike sharply. The mechanism mirrors EIP-1559 - a base fee that adjusts per block based on utilisation - but it operates on its own schedule, independent of ETH execution gas. In practice, this means a rollup’s data posting cost is now a distinct operational variable with its own volatility profile.
The Smaller Rollup Problem
For Arbitrum or Base, absorbing a temporary 10x spike in blob fees is a rounding error. For newer application-specific rollups - many of which launched post-4844 specifically because low data costs made them viable - the same spike can mean operating at a loss or delaying batch submissions. Delayed batches extend the window in which user transactions sit in a “soft finality” state rather than being anchored to Ethereum L1. That’s a security consideration most users don’t see.
Pectra, which activated in May 2025, bumped the blob count target to six with a max of nine. That added capacity has helped - sustained blob utilisation rates dropped noticeably in the months after activation - but it doesn’t eliminate the fee spike problem, it just raises the threshold before it triggers.
What Fusaka Changes
The next step is PeerDAS, expected as part of the Fusaka upgrade. The design distributes blob data across the validator network using data availability sampling rather than requiring every node to store full blob content. If it works as specified, PeerDAS could increase blob capacity dramatically without proportional increases in node storage requirements - which has been the practical ceiling on how far the Ethereum core developers are willing to push blob counts before Fusaka.
The open question is timing. Fusaka has no confirmed mainnet date as of mid-2026, and PeerDAS is technically complex enough that slippage is likely. Rollups operating today are building cost models around a blob market that may look fundamentally different in 18 months - or may not.