Before Blobs, Someone Was Paying Too Much
You're a rollup operator sitting on a batch of ten thousand user transactions. You need to post proof of that batch to Ethereum, because Ethereum's security is the whole point of using a rollup in the first place. Before EIP-4844 shipped, the only place to put that data was `calldata`: the same storage every smart contract interaction uses, priced in the same fee market, competing with every DeFi trade and NFT mint on the network. That collision was expensive. Rollup operators were spending, at times, 90% of their total costs purely on the Ethereum data-posting step.
Blob transactions changed that arithmetic. Not by magic. By introducing a second, parallel fee market with its own supply, its own price, and its own expiry schedule. The mechanism is worth understanding precisely, because it rewired the economic relationship between Layer-2 networks and Ethereum in a way that most explainers gloss over.
What a Blob Actually Is
A blob is a fixed-size chunk of data, 128 kilobytes, attached to a transaction but never executed by the Ethereum Virtual Machine. The EVM never reads it. Validators see it, attest to its availability for roughly 18 days, and then it's pruned. Gone. The on-chain footprint left behind is only a small cryptographic commitment, a KZG commitment of 48 bytes, that proves the blob existed and contained what the rollup claimed.
The critical word is availability, not storage. A rollup doesn't need Ethereum to store its transaction data forever. It needs Ethereum to guarantee the data was available long enough for anyone to reconstruct the chain state and challenge a fraudulent proof. Eighteen days is more than sufficient for that. Storing the data permanently in `calldata` was always overkill, and overkill costs money.
EIP-4844 launched with a target of three blobs per block and a maximum of six. Those numbers matter. They define blob supply, and a separate EIP-1559-style fee mechanism, with its own base fee that rises when blocks exceed the three-blob target and falls when they don't, means blob fees are almost entirely decoupled from ordinary gas fees. A congested NFT drop that spikes ETH gas prices doesn't automatically spike blob fees. The two markets breathe at different rates, like two lungs that happen to share a ribcage but not their rhythm.
The Worked Scenario: Two Rollups, One Fee Spike
Take two hypothetical rollup operators: Anya and Ben. Both post data to Ethereum. Anya runs a rollup that was live before EIP-4844. Ben launched after. During heavy mainnet congestion, a large token launch driving base fees to 150 gwei, Anya's pre-blob system is still posting `calldata`. Her batch costs roughly 4 ETH to post. Ben, using blobs, is posting the same volume of data. His blob base fee is 1 wei, because blob demand is light that day. His batch costs a few cents.
That's not a rounding error. That's a structural difference. The savings pass downstream: rollup transaction fees for users track the operator's costs over time. When data-posting costs fall 90 to 99%, user fees can fall proportionally. Optimism and Arbitrum both showed median transaction fees drop to fractions of a cent following EIP-4844's activation. The mechanism, not the moment, is what matters here.
The Part Most Guides Skip
Here's the wrinkle that deserves more attention: blob transactions didn't just lower costs, they created a new dependency. Rollups now rely on Ethereum's consensus layer to temporarily hold blob data, and they rely on a wider ecosystem of node operators and data availability samplers to actually attest that the data is there. The cryptographic commitment posted on-chain is a proof of availability, but only if the underlying blob data is actually available to be sampled. If blobs aren't being propagated correctly, the security model frays.
This is why the blob count target isn't simply set to the maximum the network can handle. Raising it too aggressively stresses the peer-to-peer network's ability to propagate large blocks quickly. The three-blob target is a conservative engineering choice, not a permanent ceiling. Future upgrades, PeerDAS and full Danksharding, are designed to push that ceiling into the hundreds of blobs per block by distributing the sampling work across more nodes. The blob transaction is, in that sense, a first chapter rather than a final answer.
What People Get Wrong About This
The common mistake is treating EIP-4844 as a Layer-2 subsidy, as if Ethereum chose to discount data costs as a favor to rollup teams. That framing misses the point entirely, and it's the wrong frame.
Ethereum's long-term scaling design has always treated rollups as the primary execution layer and Ethereum itself as a settlement and data availability layer. From that perspective, `calldata` was always the wrong tool for rollup data posting. It was a workaround. Blobs are purpose-built infrastructure, and the separate fee market isn't a discount. It's a correct pricing signal: blob data has different resource costs than EVM computation, so it should have a different price.
The folk belief that cheap blobs are a temporary gift Ethereum might take away is also worth retiring. The fee mechanism adjusts automatically. If blob demand grows enormously and blobs become expensive, that's the market working correctly, not a policy reversal.
Still, there's a real open question about what happens as blob demand scales. A world with dozens of large rollups all posting blobs regularly could saturate the current six-blob-per-block maximum. When that happens, blob fees will rise and the economics will tighten again. Whether future capacity upgrades keep pace with rollup growth is the actual variable to watch, not the current fee level.
Reading the Blob Fee Market
Any Ethereum block explorer that exposes blob data (Etherscan and Blobscan both do) will show you the blob base fee alongside the standard gas base fee. Check them both. If blob base fees are near zero, rollup operators are posting data cheaply. If they're climbing, either blob demand is rising or blob supply is being constrained by network conditions.
So here's a question worth sitting with: if you're trying to understand why rollup fees have been low, why are you still watching ETH gas fees? The blob base fee is a better leading indicator of rollup economics, and it's right there on the same page.
The deeper point: Ethereum's fee market used to be a single room where everyone competed for the same space. Blob transactions opened a second room next door, purpose-built for a specific kind of occupant. The economics of Layer-2 data availability didn't improve because rollups got lucky. They improved because someone finally built the right room, and then priced it correctly.