Blob fees were never supposed to matter. When EIP-4844 shipped with the Dencun upgrade in March 2024, the design assumption was that blob space would stay cheap and abundant - a dedicated data lane for Layer 2s that wouldn’t compete with execution gas. For roughly a year, that held. Rollup costs dropped 90% or more across major L2s almost overnight.
Then activity caught up.
By late 2025, sustained throughput from Base, Arbitrum, and a growing cluster of smaller rollups had begun pushing blob base fees into territory that made some L2 operators uncomfortable. The blob fee mechanism mirrors EIP-1559’s structure - a target and a max, with exponential repricing above the target - which means congestion doesn’t creep up linearly. It spikes. Teams running rollups on tight margins felt that acutely during high-activity periods when blob fees briefly consumed a meaningful share of sequencer revenue.
Pectra, which landed in May 2025, bumped the blob count target from 3 to 6 per block. That gave the market breathing room again. But the pattern is clear now: L2 adoption is consistently outpacing blob capacity expansion, and the question is how many upgrade cycles this can sustainably repeat.
The Rollup Count Problem
The blob market was designed with a handful of dominant rollups in mind. The actual ecosystem looks different. There are now dozens of chains posting data to Ethereum mainnet - app-specific rollups, validiums, hybrid chains - and while individually most are small, the aggregate demand adds up in ways the original blob sizing didn’t anticipate.
Ethereum’s roadmap does address this. The Fusaka upgrade, targeting late 2025 to 2026, is expected to include PeerDAS (Peer Data Availability Sampling), which would allow the network to handle significantly more blob data without proportionally burdening every node. The theoretical ceiling climbs substantially - potentially 64 blobs per block or more under full PeerDAS implementation.
What That Actually Means
PeerDAS is technically sound. But it introduces a new set of validator and client requirements, and the timeline between “spec complete” and “network stable” on major Ethereum upgrades has historically been measured in quarters, not weeks.
For now, L2s are operating in a window where blob fees are manageable but not predictable. Some teams have responded by batching more aggressively. Others are quietly evaluating alternative data availability layers - not because Ethereum DA is failing, but because fee volatility affects product planning in ways that matter at scale.
Whether that constitutes a competitive threat to Ethereum’s DA business or just healthy market pressure on the protocol to keep shipping - that’s probably the more interesting question.