You're watching a block explorer on a quiet Tuesday night. Blob base fee: 1 wei. Essentially free. Then a popular NFT mint fires up on an Optimism rollup Friday evening, and within the span of a few hundred blocks, that same slot costs orders of magnitude more. The price didn't drift upward the way gas fees do during a busy afternoon. It snapped. Getting at why requires working through the actual arithmetic of blob fee targeting, because the mechanism is genuinely unlike anything Ethereum ran before EIP-4844.

One Market, Two Speeds

Ethereum's execution gas targets 15 million gas per block, hard cap at 30 million. Blob space, introduced in EIP-4844, runs a separate, parallel market with its own numbers: target of 3 blobs per block, hard cap at 6. Those specific figures will shift as the protocol matures, but the structure won't. Two resources. Two base fees. Neither price touches the other.

The blob base fee follows the same EIP-1559 feedback loop that governs execution gas. EIP-1559 is the mechanism where the protocol automatically raises fees when blocks are full and lowers them when they're not, targeting a specific usage level. For blobs, the adjustment is 12.5% per block in either direction, identical to the gas base fee rate. Sounds familiar enough. The difference is what happens when asymmetric demand hits that 12.5% ceiling repeatedly, with nothing pushing back.

Here's the worked scenario. Call the two rollups Arca and Brev. Arca posts 5 blobs every block for an hour because its sequencer is catching up after a brief outage. Brev posts its usual 1. Total: 6 blobs per block, the hard cap. The blob base fee is now compounding at its maximum adjustment rate, up 12.5% every single block. Ethereum produces roughly 7,200 blocks a day, so 300 blocks of sustained demand means the fee has theoretically multiplied by a factor of 1.125 raised to the power of 300. That number is astronomical. In practice, demand self-limits well before that point, because rollup sequencers watch blob fees and throttle their submission rate when costs spike past their economic threshold. The market corrects. The correction mechanism is price pain, not a governor, and that distinction matters.

Flip it. Brev goes quiet for a weekend. Arca posts its normal 2 blobs. Four blocks out of every five are under-target, and the blob base fee falls 12.5% per block until it hits the protocol floor of 1 wei. That's not a metaphor or a rounding convention. A blob base fee of 1 wei is, for every practical purpose, free. Rollups get cheap data storage during quiet periods, which is exactly the design intent.

Why Asymmetry Is the Whole Point

The asymmetry that confuses people is this: blob demand doesn't come from one smooth population of users the way execution gas does. Rollup sequencers batch transactions and post in bursts. Two rollups can be completely uncorrelated, one busy because a memecoin is trending on its chain, the other idle because its primary app scheduled downtime. The blob fee market prices combined blob consumption regardless, even when the underlying demand sources have nothing to do with each other.

This is where it gets uncomfortable for rollup operators. A sequencer behaving perfectly reasonably, posting 2 blobs per block, suddenly finds itself paying elevated fees because three other rollups simultaneously hit traffic spikes and pushed aggregate consumption above target. The base fee is socialized across all blob posters proportionally. Neither Arca nor Brev caused the spike, but both pay the higher base fee if they want their data included. That feels unfair. It is, strictly speaking, unfair to the individual rollup. The protocol doesn't care, and I think that's the right call: optimizing for fairness to individual participants would make the aggregate bandwidth target unenforceable.

There's a distinction here that trips up even developers who work with this daily. The blob base fee is burned, just like the execution base fee. Rollup sequencers also attach a priority fee (a tip) to get blobs included when blocks are competitive. During quiet baselines, the base fee is negligible and tips do the sorting work. During a spike, the base fee dominates and tips become nearly irrelevant to total cost. The sequencer's economic calculation flips completely within a handful of blocks, which means any cost model built around average blob fees is about as useful as a raincoat designed for average weather.

Are you seeing your rollup's blob costs spike and crash within the same hour? That's the 12.5% adjustment mechanism doing precisely what it was designed to do. The target isn't a price anchor. It's a thermostat set to a temperature that nobody else in the building controls.

The deeper point is that blob fee targeting was never designed to be fair to any individual rollup. It's designed to keep aggregate blob usage near a level Ethereum's consensus layer can handle without straining bandwidth. From the protocol's perspective, Arca and Brev are indistinguishable sources of demand. It doesn't matter that Arca's spike came from a sequencer bug and Brev's came from genuine user activity. Both get the same repriced blob fee on the next block. Designing sequencer economics around that reality, rather than being surprised by it every time the market moves, is the actual engineering problem worth solving.