The Dencun upgrade was a coordinated change to the Ethereum blockchain that went live on 13 March 2024. Its main feature was proto-danksharding, specified in EIP-4844, which created a new, cheaper way for layer-2 networks to publish their transaction data to Ethereum in temporary packets called blobs. Fees on many layer-2 networks fell sharply in the days that followed, in some cases to fractions of a cent per transaction.
The name combines the two halves of the upgrade. Since Ethereum's switch to proof of stake in 2022, each upgrade has had an execution-layer name, taken from a city that hosted the Devcon conference, and a consensus-layer name, taken from a star. Dencun joined Cancun on the execution side with Deneb on the consensus layer. It followed the Shapella upgrade of April 2023, which enabled staking withdrawals, and was a step in Ethereum's plan to scale through rollups rather than by making the base chain itself much faster.
Dencun in plain terms
Think of Ethereum as a busy main road and layer-2 networks as delivery companies that bundle many parcels into one lorry. Before Dencun, each lorry had to pay to park permanently on the main road, because the companies stored their delivery records there forever. That was expensive, and the cost was passed on to customers.
Dencun opened a separate, cheaper car park just for those records. Records left there stay long enough for anyone to check them, about two and a half weeks, and are then cleared away. Because the delivery companies no longer pay for permanent space, their costs, and the fees their customers pay, dropped a great deal.
What changed
The upgrade bundled several Ethereum Improvement Proposals, each described in the EIP process. The most important were:
- EIP-4844 (proto-danksharding). Added blob-carrying transactions, a new transaction type that attaches large data packets to blocks without making them permanently readable by smart contracts.
- EIP-1153. Introduced transient storage, a cheap form of storage that is cleared at the end of each transaction and is useful for complex smart contracts.
- EIP-4788. Made recent consensus-layer block roots available inside the Ethereum Virtual Machine, helping staking and bridge applications verify information about the Beacon Chain.
- EIP-6780. Restricted the old SELFDESTRUCT instruction, removing a source of complexity and risk.
- EIP-7514. Limited how many new validators could join per epoch, slowing the growth of the validator set.
Like every Ethereum upgrade, Dencun was a hard fork: all node operators had to install updated software before the activation time, and nodes that did not upgrade would have fallen off the main network. Before reaching mainnet it was rehearsed on public test networks, including Goerli, Sepolia and Holesky, during January and February 2024, and the activation went through without a significant disruption. The upgrade took effect at a pre-announced epoch, the unit of time used by the consensus layer.
How blobs work
Each blob holds about 128 kilobytes of data. At launch the network targeted three blobs per block, with a maximum of six. Blob space has its own fee market, separate from ordinary gas fees and modelled on the base-fee mechanism introduced by EIP-1559: when blocks carry more than the target number of blobs, the blob fee rises; when they carry fewer, it falls. Consensus nodes keep blob data for roughly 18 days and then delete it. That window is long enough for anyone to check the data and challenge an invalid rollup state, while preventing the permanent growth of Ethereum's history that storing the same data as ordinary calldata would have caused.
The "proto" in proto-danksharding reflects that EIP-4844 put the format and fee market in place without yet splitting data across nodes. Full danksharding is meant to add data availability sampling, letting nodes verify large amounts of blob data by checking small random pieces, and thereby to increase blob capacity many times over, a goal related to the older idea of sharding.
Effects
The main beneficiaries were rollups. Networks using optimistic rollups, such as Arbitrum, and zero-knowledge rollups switched to posting data in blobs and passed most of the savings on to users. Lower costs encouraged a surge of activity on layer-2 networks, which by late 2024 together processed many more transactions than Ethereum's base layer. Blob space was underused at first, so blob fees often sat near their minimum; as demand grew, a later upgrade, Pectra, in May 2025, raised the target to six blobs and the maximum to nine.
The upgrade also prompted debate. Because rollups now paid Ethereum far less for data, the amount of ether burned through fees fell, and some commentators argued that layer-2 networks were capturing value at the base layer's expense. Supporters replied that cheap data was the intended outcome and that fees would rise as blob demand grew. Dencun did not change fees for ordinary transactions made directly on Ethereum's main chain.