Danksharding

Layer 2 Updated May 2026

What is Danksharding?

Danksharding is Ethereum’s full scalability roadmap — the eventual end state where Layer 2 rollups can achieve 100,000+ transactions per second. It’s named after Dankrad Feist, an Ethereum researcher who proposed the architecture.

Danksharding builds on EIP-4844 (Proto-Danksharding), which is already live. While EIP-4844 introduced blobs with limited capacity, full danksharding massively scales blob capacity using data availability sampling (DAS).

The Two Phases

Phase 1: EIP-4844 (Proto-Danksharding) — LIVE since March 2024
  → ~6 blobs per block (~128 KB)
  → No data availability sampling
  → ~100-500 tx/s on L2s

Phase 2: Full Danksharding — Future
  → ~64 blobs per block (~2 MB+)
  → Data availability sampling (DAS)
  → Peer-to-peer sampling
  → ~100,000+ tx/s on L2s

The Key Innovation: DAS

Data Availability Sampling (DAS) is what makes danksharding work. The idea is elegant:

The Problem

To ensure rollup data is available, nodes traditionally must download ALL of it. At scale (2 MB per block), this becomes impractical for most users.

The Solution

Instead of downloading everything, each light node downloads small random chunks of the data:

Block data: [A][B][C][D][E][F][G][H][I][J][K][L]...[Z]

Node 1 samples: [C][J][Q]     → "My chunks are here ✓"
Node 2 samples: [A][K][W]     → "My chunks are here ✓"
Node 3 samples: [D][F][Z]     → "My chunks are here ✓"
...
Thousands of nodes collectively verify all chunks

If the data publisher withholds ANY chunk:
  → Some node will sample the missing chunk
  → Network detects withholding
  → Block is rejected

With enough nodes sampling random chunks, the probability that a withheld chunk goes undetected approaches zero. This lets the network support massive blob sizes without requiring every node to download everything.

PeerDAS: The Stepping Stone

Before full danksharding, Ethereum will implement PeerDAS (Peer-to-peer Data Availability Sampling):

Each node downloads only a subset of blob columns
Nodes share data through a peer-to-peer gossip network
Reduces individual node bandwidth requirements
Expected to increase blob capacity to ~8–16 blobs per block

PeerDAS is the next incremental step toward full danksharding.

danksharding’s Other Components

Beyond DAS, full danksharding includes:

Component	Purpose
2D Reed-Solomon Erasure Coding	Extends data so partial chunks can reconstruct the whole
Builder-Proposer Separation (PBS)	Separates who builds a block from who proposes it, reducing MEV centralization
Censorship Resistance CrLists	Lets proposers force inclusion of transactions builders might censor
Proposer-Builder Separation	Enshrined in protocol to prevent builder centralization

Impact on L2 Ecosystem

Metric	EIP-4844 (Now)	Full Danksharding (Future)
Blob space per block	~128 KB	~2 MB
L2 fees	$0.01–$0.10	<$0.001
L2 throughput	~100–500 tx/s	~100,000+ tx/s
Light node requirements	Full blob download	Sampled chunks only

At full danksharding, L2 fees could drop below $0.001 — making crypto transactions competitive with traditional payment rails.

Timeline

Danksharding is a multi-year roadmap:

Milestone	Status	Expected
EIP-4844 (Proto-Danksharding)	✅ Live (March 2024)	Done
PeerDAS	In development	2025–2026
Full DAS	Research phase	2026+
Full Danksharding	Research phase	2027+

These dates are estimates — Ethereum upgrades follow a conservative, research-driven timeline.

Frequently Asked Questions

Q: Is danksharding the same as sharding? A: No. Original Ethereum sharding split the chain into multiple shards (chains). Danksharding shards data, not execution. All execution stays on L1/L2 — only the data availability layer is “sharded.”

Q: How is danksharding different from Celestia? A: Celestia is a separate DA blockchain. Danksharding brings DA scaling to Ethereum itself. Both use similar techniques (DAS, erasure coding).

Q: Will danksharding make L1 transactions cheaper? A: No. Danksharding only scales L2 data availability. L1 transaction costs remain governed by regular gas economics.