Full Danksharding: The Final Shard Architecture

Rollups post data to Ethereum for security, but L1 block space is finite and expensive. This creates a scalability ceiling and keeps transaction fees high for end-users, even on L2s.

• Cost Bottleneck: >90% of a rollup's operational cost is L1 data posting.
• Throughput Wall: Current calldata limits cap total network throughput at ~100 TPS across all rollups.

Introduces blob-carrying transactions, a dedicated data channel for rollups separate from execution. This is a prerequisite for Full Danksharding, delivering an immediate 10-100x cost reduction for L2s.

• Blob Space: ~1.3 MB per block of cheap, ephemeral data.
• Decoupling: Separates data pricing from EVM gas, creating a commoditized data market.

Scales data availability to ~1.3 MB per shard, per slot, with 64 shards planned. Validators only sample small pieces of data via Data Availability Sampling (DAS), making verification trustless and lightweight.

• Throughput: Enables ~100,000 TPS across the L2 ecosystem.
• Security: Inherits Ethereum's consensus; no new trust assumptions.
• Efficiency: KZG commitments and erasure coding ensure data is available and recoverable.

Unlike earlier sharding proposals (e.g., execution shards), Danksharding only shards data availability. Execution remains the domain of rollups like Arbitrum, Optimism, and zkSync. This simplifies consensus and leverages L2 innovation.

• Validator Simplicity: Nodes verify data availability with ~50 KB downloads per slot.
• Modular Triumph: Cements Ethereum's modular stack: L1 for consensus/DA, L2 for execution.

Proto-Danksharding (EIP-4844) introduced blob-carrying transactions with a ~18-day expiry, creating a separate fee market. This is a stepping stone, not the final state. Rollups still compete for limited blob space (~3-6 per block), and nodes must store this data temporarily, missing the full scaling benefits.

• Interim Design: Blobs are a separate sidecar, not integrated into execution.
• Storage Overhead: Nodes store blobs for weeks, not seconds.
• Capacity Ceiling: ~0.375 MB per block is a soft limit, not the target.

Full Danksharding transforms the Beacon Chain into a pure data availability (DA) engine. It implements Data Availability Sampling (DAS), where light clients can cryptographically verify data availability by sampling tiny random chunks. This enables the target of ~128 blobs (~16 MB) per slot.

• DAS Magic: Enables secure scaling; you don't need to download all data to know it's there.
• No More Expiry: Data becomes part of the canonical chain history.
• Execution Separation: Validators and builders handle DA; execution clients (like Geth) only process small commitments.

Two cryptographic primitives make this trust-minimized scaling possible. KZG polynomial commitments create small, efficiently verifiable proofs that the sampled data is correct. Proofs of Custody force validators to cryptographically prove they actually possess the blob data they attest to, slashing those who lie.

• Trustless Sampling: KZGs allow verification without full data download.
• Validator Accountability: Proofs of Custody eliminate data withholding attacks.
• Efficiency: A single KZG commitment can represent megabytes of data.

The core innovation is arranging blob data in a two-dimensional Reed-Solomon erasure-coded matrix. If up to 50% of the data is missing, it can be fully reconstructed from the remaining samples. This redundancy is what makes Data Availability Sampling both secure and robust.

• Redundancy: Data is encoded with 2x redundancy (extending rows and columns).
• Robust Recovery: Network can tolerate significant data loss.
• Sampling Simplicity: Clients randomly sample a handful of points from this 2D grid.

With abundant, cheap, and permanent DA on Ethereum, rollups (like Arbitrum, Optimism, zkSync) transition from cost-constrained tenants to sovereign execution layers. Their cost structure flips from variable, congested L1 gas to a minimal, predictable DA fee. This enables micro-transactions, full data on-chain, and complex state transitions.

• Cost Predictability: L2 tx costs dominated by fixed DA posting, not volatile execution gas.
• Design Freedom: Rollups can afford to post all data, enabling easier fraud proofs and interoperability.
• Ultimate Goal: Enables ~100k TPS across the rollup ecosystem.

Full Danksharding is a multi-year roadmap dependent on prior upgrades. Key prerequisites include PeerDAS (distributed sampling), Verkle Trees for statelessness, and a robust builder market. The main hurdles are implementing complex P2P networks for sampling and ensuring validator hardware can handle the load.

• Prerequisite Chain: PeerDAS -> Verkle Trees -> Full Danksharding.
• Hardware Load: Validators need sufficient bandwidth and SSD speed for sampling.
• No Magic Date: This is Ethereum's final major scaling upgrade, likely post-2025.