Glossary
Sharding (blockchain)
Sharding splits a blockchain's work across parallel parts to raise capacity; Ethereum replaced it with danksharding, starting with blobs in March 2024.
On this page
Sharding splits a blockchain into parts, called shards, that process transactions in parallel, so that no single node has to handle all of the network's work. The term comes from databases, where a large table is split across servers. The aim is more capacity without requiring every node to run bigger hardware.
Ethereum's change of plan
Ethereum's original roadmap called for 64 shard chains, each running its own transactions. Moving contracts and balances safely between shards proved hard. Meanwhile layer-2 rollups had begun doing the execution off the main chain, posting compressed transaction data back to it. Ethereum dropped execution shard chains from its roadmap and turned to danksharding: the main chain stays one chain, and the scaling goes into cheap space for rollup data.
The first step, proto-danksharding (EIP-4844), went live in March 2024 in the Dencun upgrade. It added blobs: 128 KB data packets attached to blocks, priced in their own fee market and deleted by nodes after about 18 days. The main chain cannot read a blob's contents; it only keeps a commitment to them.
The next step relies on data availability sampling, in which each node checks small random pieces of the blob data instead of downloading all of it. The Fusaka upgrade of December 3, 2025 switched it on (PeerDAS, EIP-7594), and two follow-up forks raised the blob target and maximum to 10 and 15, then to 14 and 21 in January 2026. Full danksharding aims to go further, to 64 blobs per block.
Elsewhere
Some chains shard from the start. NEAR Protocol splits its state and transaction processing across shards under a design it calls Nightshade. Sharding trades simplicity for capacity: cross-shard transactions are slower and more complex, and each shard is secured by a smaller share of the network.
