PeerDAS took Ethereum's blob target from six to 14 in 35 days on validators' existing hardware

Crypto Briefing's "more than 95%" depends on which ends of two ranges you pair ^[6]. At the midpoints, a $0.35 layer 2 transaction became a $0.0125 one, a cut of about 96% ^[1]. Pair the cheapest fee from before the upgrade with the most expensive one after it, $0.20 against $0.02, and the cut is 90% ^[2]. The opposite pairing, $0.50 against half a cent, gives 99% ^[3].

Fees fell much further than capacity rose. The blob target went from six to 14, about 2.3 times ^[4], while the midpoint cost of a transaction fell by a factor of 28 ^[5]. Crypto Briefing says the reduction is not purely a function of PeerDAS ^[7]. EIP-4844 had already introduced blob-carrying transactions in early 2024 and separated blob fees from execution gas fees. PeerDAS builds on that by letting far more blob data through without a matching rise in each validator's burden ^[7].

For a rollup, the user fee is its price and blob space is one of its costs. The article reports user fees only. It does not say what rollups pay Ethereum for blob space or how much of the saving they kept. A 96% fall in what users pay ^[1] fits two different stories. In one, rollups passed the cheaper data straight through to users and kept the same margin per transaction. In the other, they also cut their own markup to compete, so their margins fell along with the fee.

The hardware claim rests on the design and one test run. Before mainnet activation, developers ran the protocol across hundreds of nodes on a dedicated devnet ^[9]. Researchers including Danny Ryan and Dankrad Feist spent years on the theory before implementation began ^[10]. PeerDAS is an early phase of Feist's larger danksharding design ^[8].

The picture could look different in a year for any of three reasons. Blob demand could grow into the 14-blob target and push fees back toward where they started. Later increases, on a path the article says could reach thousands of blobs per block under full danksharding ^[8], could need hardware that the first 14 blobs did not. Or the fee record could owe more to EIP-4844 than to PeerDAS ^[7]. I think the evidence supports a narrower claim. Ethereum more than doubled its blob target within 35 days of launch ^[7], and by Crypto Briefing's account the protocol has run for nearly a year without a major incident ^[2]. The thesis is wrong if a later increase causes an incident or forces validators to buy new equipment.