What Is Sharding in Blockchain? Meaning and Examples

Sharding is a blockchain scaling technique that splits a network into smaller groups, called shards, that process transactions and store data in parallel instead of every node doing all the work. It is one of the main ways a layer-1 blockchain can raise throughput without asking every validator to run bigger and bigger machines.

Sharding definition

The term comes from databases, where a large table is split into horizontal "shards" that live on different servers. Applied to a blockchain, sharding means each validator only processes and stores part of the network's activity, while the system stays one chain with one token.

There are three kinds of sharding, and a network can use one or all of them:

• Network sharding: validators are split into groups, each responsible for one shard.
• Transaction sharding: each transaction is processed by one shard rather than by every node.
• State sharding: each shard stores only its part of the ledger (balances, contract storage), so no single node has to hold everything. This is the hardest kind to build, because transactions often touch accounts on different shards.

How sharding works

A state-sharded chain typically works like this:

1. Account assignment. Every account and smart contract lives on one shard, usually determined by its address.
2. Validator assignment. Validators are randomly assigned to shards and periodically reshuffled, so attackers cannot predict or concentrate on one shard.
3. Parallel blocks. Each shard produces its own blocks at the same time as the others.
4. Coordination. A coordinating chain (MultiversX calls it the metachain) notarises shard block headers so the shards agree on one global history.
5. Cross-shard messages. When a transaction sends value or calls a contract on another shard, it executes in two steps: debit on the sender's shard, then credit or execution on the receiver's shard. That makes cross-shard calls asynchronous.

Approaches differ by chain:

• MultiversX uses Adaptive State Sharding, combining network, transaction and state sharding. The network config returned 3 execution shards plus the metachain when we checked on 10 October 2026, and the design allows shards to be added or merged as load changes (MultiversX docs). Its Supernova upgrade was scheduled for epoch 2233 (10 September 2026) to cut block time from 6 seconds to 600 ms. The live config showed a 600 ms round duration at epoch 2259 on 10 October 2026, so the upgrade is active.
• Ethereum dropped its original plan to shard execution and moved to a rollup-centric roadmap called danksharding, which shards data availability for layer-2s rather than execution. Proto-danksharding (EIP-4844, "blobs") shipped in the Dencun upgrade on 13 March 2024. PeerDAS (EIP-7594), which lets nodes sample only part of the blob data, followed in the Fusaka upgrade in December 2025 (ethereum.org, Blockworks).
• Sui does not split state into shards. It scales through an object-centric data model: transactions that touch only owned objects can skip consensus, transactions on shared objects are ordered by the Mysticeti consensus protocol, and transactions with no overlapping objects execute in parallel (Sui docs).

Sharding example

Hypothetical, using round numbers: say a non-sharded chain processes 1,000 transactions per second, limited by what one validator can execute. Split it into 4 shards and, if every transaction stays inside one shard, the network can in theory handle about 4 × 1,000 = 4,000 transactions per second, because each shard works in parallel.

Real traffic is messier. Say 30% of transactions are cross-shard and each one costs roughly double the work, because it is processed once on each side. Average work per transaction is then 0.7 × 1 + 0.3 × 2 = 1.3 units, so effective capacity is about 4,000 ÷ 1.3 ≈ 3,077 transactions per second. Still a gain, but below the headline number.

Why sharding matters

• Lower fees under load. More capacity means less competition for block space, which keeps gas fees lower during busy periods.
• Developer complexity. Asynchronous cross-shard calls mean a contract cannot always read another contract's state and act on it in the same transaction. Apps have to handle callbacks and partial failures.
• Security trade-offs. Each shard is secured by a subset of validators, so random assignment and frequent reshuffling are essential. Sharding also ties into the network's consensus design; see proof of stake vs proof of work.

Sharding on JewelSwap

JewelSwap's MultiversX smart contracts, like every MultiversX account, each live on a specific shard. When your wallet is on a different shard from a contract, your transaction is processed as a cross-shard call: it leaves your shard first and executes on the contract's shard shortly after, and contract-to-contract calls across shards are asynchronous. This is normal behaviour on MultiversX, not an error. Our guide to JewelSwap on MultiversX covers what runs there, and our list of the best DeFi protocols on MultiversX shows the wider ecosystem.

Related terms

• Gas fees — the cost of getting a transaction included on-chain.
• Epoch — a fixed period after which validators, including shard assignments, can rotate.
• Slashing — penalties for validators that misbehave or go offline.
• Crypto bridge — moving assets between separate chains, unlike cross-shard transfers within one chain.
• LST (liquid staking token) — a token representing staked assets on a proof-of-stake chain.

Learn more on the JewelSwap blog

• Proof of stake vs proof of work — the consensus models that sharded chains are built on.
• Best DeFi protocols on MultiversX in 2026 — what is being built on the best-known sharded layer-1.
• JewelSwap on MultiversX — JewelSwap's products on MultiversX and how to use them.
• What are gas fees? — why block space costs money and how network capacity affects it.

Frequently asked questions

What is sharding in simple terms?

Sharding splits a blockchain into smaller parts, called shards, that process transactions in parallel. Each validator handles only one part, so the network as a whole can process more transactions than a single machine could.

What is adaptive state sharding?

Adaptive State Sharding is MultiversX's design. It combines network, transaction and state sharding, uses a metachain to coordinate the shards and is built so shards can be added or merged as demand changes.

Sharding vs layer-2 rollups: what is the difference?

Sharding scales the base layer itself by splitting it into parallel shards. Rollups execute transactions on a separate layer-2 and post data back to the base chain. Ethereum combines the two ideas: danksharding shards data availability to make rollups cheaper.

JewelSwap Crypto Glossary · educational, not financial advice. Updated 10 October 2026. Browse the full glossary.