Ethereum's Dencun Upgrade:
What EIP-4844 Proto-Danksharding Means for Layer 2 Fees

On March 13, 2024, Ethereum activated its most consequential upgrade since the Merge: Dencun. Named after the combination of two consensus-layer and execution-layer hard forks (Deneb + Cancun), this upgrade shipped nine Ethereum Improvement Proposals — but one of them, EIP-4844, stole the show entirely.

The result was immediate and dramatic. Within hours of activation, transaction fees on Arbitrum, Optimism, Base, and zkSync dropped between 80% and 95%. What once cost users $0.30–$2.00 per swap suddenly cost fractions of a cent. This article explains why, and what it means for the long-term trajectory of Ethereum and its Layer 2 ecosystem.

The Problem: Why L2 Fees Were So High Before

Layer 2 networks process transactions off-chain, but to remain trustless and secure, they must periodically publish compressed transaction data back to Ethereum Layer 1. This is the entire security model: if an L2 ever misbehaves, anyone can reconstruct the state from the data on Ethereum and submit a fraud proof or validity proof.

Before Dencun, this data was posted as calldata — a field in a regular Ethereum transaction that is stored permanently in Ethereum's history. Calldata is expensive because it's part of Ethereum's permanent state, and every full node must store it forever.

The cost formula was brutal: even compressed, a single L2 batch posting might use 100,000+ bytes of calldata at 16 gas per non-zero byte. With Ethereum gas at 30–50 gwei, a single batch cost $50–$200 — and that cost was spread across every user in the batch, driving up individual fees significantly.

The Solution: Binary Large Objects (Blobs)

EIP-4844 introduces a brand-new transaction type — Type 3 — that carries one or more "blobs" attached to a transaction. Each blob is exactly 128 KB of data, and each Ethereum block can contain a target of 3 blobs (max 6), giving a target throughput of ~384 KB per block.

The key innovation: blobs are not stored in Ethereum's execution layer state. Full nodes store them for approximately 18 days (4,096 epochs), after which they can be pruned. This means:

• Blob data does not bloat Ethereum's permanent history
• All validator nodes still verify blob availability at the time of the block
• L2s can reconstruct any disputed state within the 18-day window (more than enough for challenge periods)
• A separate blob gas market keeps blob fees extremely low and independent from regular gas fees

Real-World Fee Reduction: The Numbers

These fee reductions aren't just quality-of-life improvements — they are fundamental to Ethereum's ability to compete with Solana, BNB Chain, and other monolithic blockchains for everyday user activity. Sub-cent transactions make DeFi, NFT trading, and on-chain gaming economically viable for retail users worldwide.

How Blob Commitments Work: KZG Proofs

One elegant aspect of EIP-4844 is that while blob data itself is transient, a cryptographic commitment to each blob is stored permanently in the execution layer. This commitment uses KZG polynomial commitments — the same cryptography that powers many ZK proof systems.

The commitment scheme allows anyone to prove that a specific piece of data was included in a specific blob at a specific block height, without needing the full blob. This is crucial for ZK rollups that use the blob data as a data availability layer while generating proofs later.

The Road to Full Danksharding

EIP-4844 is explicitly named "proto-danksharding" because it lays the groundwork for the full Danksharding vision without implementing all of it. Currently, blob availability is verified by all full nodes downloading the blob data — a straightforward but bandwidth-intensive approach.

Full Danksharding will introduce Data Availability Sampling (DAS), where each node only downloads a small random sample of each blob, yet can statistically guarantee — with overwhelming probability — that all blob data is available. This will allow the blob capacity to scale to hundreds of blobs per block, pushing Ethereum's data throughput into the gigabyte-per-block territory.

Proto-danksharding is the most important step towards a scalable Ethereum we've taken since the Merge. Fees going from dollars to cents is not a minor UX improvement — it's the difference between a global financial system and a niche product for wealthy users.

— Vitalik Buterin, ETHDenver 2024

What This Means for L2 Developers and Users

For developers building on L2 networks, Dencun fundamentally changes economic assumptions. Applications that were previously unviable — micro-transactions, on-chain gaming state, frequent rebalancing strategies — are now economically practical.

For users, the practical takeaway is straightforward:

• DeFi swaps on any major L2 now cost $0.01–$0.05, making frequent rebalancing sensible
• NFT minting on L2s is now comparable to centralized platforms in cost
• Cross-chain bridging is cheaper as bridge contracts batch more efficiently
• On-chain gaming with frequent small transactions is finally viable at scale

The Bottom Line

Dencun's EIP-4844 is the single most impactful Ethereum upgrade for everyday users since the London hard fork's EIP-1559 introduced fee burning. It validates the rollup-centric roadmap Ethereum has pursued since 2020 and demonstrates that the multi-layer approach to scaling is paying off in concrete, measurable ways.

The next milestone — full Danksharding with DAS — is still years away, but the path is now clearly laid out. In the meantime, Ethereum's Layer 2 ecosystem is positioned to handle the next wave of onboarding with fees that can compete with any blockchain on the market.