What is a Shred?
A shred is the most basic unit of block data on Solana. It’s a piece of a block that has been sized to fit into a single network packet (approximately 1,200 bytes). The leader does not send blocks anywhere when it has produced a block. Rather, a leader produces and begins sending blocks to the network while it is still being built. The rest of the network begins assembling the block before the block is built.
Think about sending a book page by page instead of once it’s been fully written. Your readers will be able to read parts of the book while the author is still writing other parts.
How do shreds work?
There are two types of shreds. Data shreds, like what we’ve discussed, are data packets containing transaction info. Coding shreds are shreds that are used to reconstruct the full block using erasure coding. Erasure coding is a technique that allows the recipient to reconstruct missing data by combining it with data already received. Shreds are sent using a UDP connection. This means they are not guaranteed to arrive. If any data packets are missing from the original block, the recipients can use erasure coding to reconstruct the full block based on the parts they have.
Each shred is created by the leader that was selected for the current block time. That shred is given a timestamp, a slot number, and a shred number (the specific position that the shred is in the full block), and is then broadcast through the Turbine consensus tree (a stake-weighted dissemination tree). Recipient nodes collect the shreds and reassemble them back into a block, then replay the block’s transactions and vote on their validity. In a feature slated to be implemented in a 2026 upgrade, shreds will instead be routed through something called Rotor, an update to Turbine that eliminates multiple-hop propagation by relaying shreds to all nodes directly from the leader. But, shreds themselves are still the core unit of a block.
Why Shreds Are Valuable Beyond Consensus
It turns out shreds have value beyond just being used for consensus. If you receive transaction data from shreds before the rest of the network, you have the ability to react much faster. And in markets, this can mean a significant advantage. A validator receives transaction data before the transaction has reached consensus status, potentially hundreds of milliseconds before the full block arrives at validators. This is a significant edge in a trading environment.
An ecosystem has emerged around the value of shreds, with validators who run their own shred streams and charge fees to subscribe to their data. For example, jito’s ShredStream delivers raw shreds to any node with minimal latency. In Q2 of 2026, DoubleZero Edge, a network running on a dedicated fiber network built by DoubleZero, had 434 validators using it to publish shreds and traders using a USD-backed stablecoin known as USDC to pay for subscription access to the stream. In some cases, when a token starts trading on Raydium, bots racing to buy tokens before others are racing based on what they see in shreds, not necessarily on confirmed blocks.
Shred data should not be used for any critical use-cases because it is unconfirmed. You get access to it before consensus, meaning that there is a risk that it will be rejected and removed from the blockchain, even as it’s being received by the validators you’ve been connected to.
Why Shreds Matter
Blocks on other chains, like Ethereum and Bitcoin, are only propagated after they’ve been built. In Ethereum’s case, a block is built and then gossiped across the network in the 12-second timeframe for each slot. In Solana, though, it’s the ability to send blocks before they’ve been fully built that makes a 400ms slot time possible. Blocks are propagated at the same time they’re being built, instead of waiting until they’ve finished being built. That means you can confirm transactions on Solana on your wallet (I use a Phantom wallet) in about a second for a base transaction cost that was around $0.00025 at the time of writing.
That’s the value of shreds to Solana. They allow for block data to be propagated to the network in packets that can be reconstructed via erasure coding, which allows 1,000+ validators across the globe to keep up with a 400ms block clock. That means users get confirmation in a second, rather than a minute or longer. That means faster, cheaper transactions for users and more up to the minute data for developers and traders to access. And that means the ability to build networks that don’t suffer from the inherent trade-offs of gossip-based block propagation that most other major blockchains accept.
