Ethereum Data Availability for Rollups: Blobs to DAS
Ethereum provides data availability for rollups through blob transactions, ensuring anyone can verify off-chain activity without trusting the sequencer.
Key takeaways
- Ethereum's role in the rollup stack is data availability and settlement
- Data availability means ensuring transaction data remains verifiable long enough for proofs to settle, not permanent storage
- EIP-4844 (Proto-Danksharding) fundamentally separated DA from execution by introducing a dedicated blob lane with its own fee market
- KZG commitments are the cryptographic backbone that allow Ethereum to verify blob availability without every node downloading full data
Ethereum serves as the data availability layer for rollups by temporarily storing compressed transaction data – called blobs – so that anyone can verify the rollup's state without trusting its operator. This guarantee is what gives rollup security its teeth.
Learning about how Ethereum provides this guarantee and where it is headed next is essential for anyone building on or evaluating the L2 ecosystem in 2026.
Why Rollups Need Ethereum as Their DA Layer
| Quick answer: Rollups need a DA layer because execution alone does not guarantee security. The underlying transaction data must be publicly verifiable. Ethereum is the default choice because it combines settlement and DA on the same chain, eliminating any additional trust assumptions. |
When a rollup processes transactions, it batches them off-chain and posts a summary back to Ethereum. But a summary alone is not enough. The raw transaction data must also be available so that any third party can re-execute the batch and catch a fraudulent state update. In an optimistic rollup, this data is what enables fraud proofs. In a ZK rollup, it is what allows anyone to reconstruct state independently of the prover.
If a sequencer withholds that data, the fraud proof window passes without challenge, and invalid state transitions can be finalized. The rollup's security model collapses because the underlying data simply was not there.
This is why rollups need a DA layer, and why Ethereum is the natural default:
- Settlement and DA live on the same chain, eliminating the need for additional trust assumptions
- Ethereum's validator set is the most decentralized and economically secure available
- Any rollup using Ethereum for DA inherits Ethereum's full security model without cross-chain dependencies
DA does not mean permanent storage. It means that data is available long enough for the relevant proof window to close, typically around 18 days for current rollup designs. After that, the data can be safely discarded.
The Calldata Problem: Why Ethereum DA Was Broken Before EIP-4844
| In short: Before March 2024, Ethereum had no dedicated mechanism for rollup data. Rollups had to use calldata – the same data field used to pass input to smart contract calls. |
This created a structural mismatch in three ways:
1. Calldata was priced for permanence, not verification
Calldata is stored permanently as part of Ethereum's execution layer history. Every Ethereum full node keeps it indefinitely. But rollups only need their data available for ~18 days. They were paying for permanent storage they did not need.
2. Calldata competed in the wrong fee market
Rollup data and smart contract execution shared the same gas market. During periods of high on-chain activity, like DeFi surges or NFT mints, calldata costs spiked alongside execution fees, even though the rollup's data demand had nothing to do with those events.
3. The cost burden fell almost entirely on users
Over 90% of a rollup user's transaction fee was attributable to the cost of posting calldata to Ethereum L1 (Ethereum.org). Rollups were cheap at the execution layer but expensive at the data layer – the exact opposite of what a scaling solution should look like.
By late 2023, major rollups like Arbitrum and Optimism were collectively spending tens of thousands of dollars per day on calldata alone (arXiv, 2411.03892). The architecture was not sustainable.
How Ethereum's EIP-4844 Created a Dedicated DA Lane
| In short: EIP-4844, activated with the Dencun upgrade on March 13, 2024, addressed the calldata problem directly by introducing a new transaction type built specifically for rollup data: the blob-carrying transaction. |
The core design principle is separation. Blob data is not part of Ethereum's execution layer. It lives in the consensus layer (Beacon Chain), it has its own fee market, and it is automatically deleted after approximately 18 days. The EVM cannot read blob contents at all. It can only access a short cryptographic reference called a KZG commitment.
How blobs work in practice:
- A rollup batches its off-chain transactions and compresses them into blob data
- The rollup submits a blob-carrying transaction to Ethereum, attaching the blob and its KZG commitment
- Consensus clients verify the commitment and store the blob for ~18 days
- Any node or user can download the blob during this window to verify the rollup's state
- After the retention period, the blob is pruned. Only the KZG commitment remains on-chain permanently
What KZG commitments do:
A KZG (Kate-Zaverucha-Goldberg) commitment is a polynomial commitment scheme that allows Ethereum to prove a blob was available without every node needing to download the entire blob. Think of it as a compact, verifiable fingerprint of the data. If the data matches the commitment, it is genuine. This is the cryptographic foundation that makes blob-based DA trustworthy at scale.
The fee market impact:
Blob gas operates as a completely separate market from execution gas, using an EIP-1559-style adjustment mechanism. This means blob fees respond to rollup DA demand, not to DeFi activity or NFT minting. The result was immediate: L2 data posting costs dropped by approximately 90% on the day Dencun activated (Eco.com, May 2026).
Calldata | Blob (EIP-4844) | |
| Storage duration | Permanent | ~18 days |
| Storage layer | Execution layer | Consensus layer (Beacon Chain) |
| EVM access | Full read access | KZG commitment reference only |
| Fee market | Shared with execution gas | Separate blob gas market |
| Approximate cost | ~$100/MB | ~$3–4/MB (at launch) |
What Data Do Rollups Publish to Ethereum?
The type of data a rollup posts to Ethereum depends on its proof model.
- Optimistic rollups (Arbitrum, Optimism, Base) post full transaction data. Anyone needs to be able to re-execute every transaction to submit a valid fraud proof during the challenge window (~7 days). Without full data, the fraud proof mechanism cannot function.
- ZK rollups (zkSync, Starknet, Scroll) post state diffs – the differences between the pre- and post-batch state – rather than full transaction data. Because validity proofs cryptographically guarantee correctness, users do not need to re-execute transactions. They only need state diffs to reconstruct the current state if needed.
In both cases, the data must be verifiable by anyone, not just the rollup's own nodes. This is the non-negotiable requirement that Ethereum's DA layer satisfies.
Ethereum DA vs Modular DA: Which Is Better for Rollups?
| Quick answer: Ethereum native DA offers the strongest security guarantees with no additional trust assumptions, but at a higher cost. Modular DA layers like Celestia and EigenDA are significantly cheaper, but introduce separate trust dependencies that rollups and their users must accept. |
Since EIP-4844, an ecosystem of dedicated DA layers has emerged as a direct alternative to Ethereum native DA. The most significant are Celestia, EigenDA, and Avail. Each takes a different approach:
Ethereum Native DA
- Trust model: no additional assumptions beyond Ethereum itself
- Settlement and DA live on the same chain
- Cost (mid-2026): ~$3–4/MB for blob space
- Security: backed by Ethereum's full validator set and economic security
Celestia
- Purpose-built DA chain using Data Availability Sampling (DAS)
- Cost: approximately $0.07/MB, roughly 55x cheaper than Ethereum blobs at equivalent periods (Eclipse Labs, April 2025)
- Trade-off: introduces a separate trust assumption on Celestia's validator set (~100 validators); uses fraud proofs, meaning full DA finality has a ~10-minute challenge window
- Well-suited for: high-throughput use cases, gaming L3s, sovereign rollups
EigenDA
- Built on EigenLayer; leverages ETH restaking rather than a separate validator set
- Throughput: ~15 MB/s, significantly higher than current Ethereum blob capacity
- Cost: substantially cheaper than Ethereum blobs for high-volume data posting
- Trade-off: relies on a Data Availability Committee (DAC) model. Economic security is strong, but it is not publicly verifiable DA in the same sense as Ethereum or Celestia
- Well-suited for: large Ethereum-native L2s that want alignment with Ethereum security assumptions without paying native DA prices
The honest trade-off:
Ethereum DA | Celestia | EigenDA | |
| Trust model | Ethereum only | Ethereum + Celestia validators | Ethereum + restakers |
| DA finality | ~12–15 min (Ethereum) | ~10 min challenge period | ~12–15 min (inherited) |
| Cost | Higher | Lowest | Low |
| Security alignment | Maximum | Independent chain | ETH-aligned |
As of 2026, DA costs are the single largest variable cost for most rollups, making DA choice one of the most consequential architectural decisions a rollup team makes. There is no right answer without first defining the rollup's security requirements, user base, and cost targets.
What Ethereum's DA Roadmap Means for Rollup Builders
| In short: Ethereum's DA roadmap means that the cost gap between native DA and modular alternatives will narrow over time, making today's choice between cost and security a temporary trade-off, not a permanent one. |
Stage 1: EIP-4844 (Live, March 2024)
Introduced blobs and a separate DA fee market. Blob target: 3 per block, maximum: 6 per block. Post-Pectra (May 2025), capacity was doubled, and BPO (Blob Parameter Only) forks pushed the target to 14 and maximum to 21 per block by mid-2026.
The combined L2 ecosystem now processes approximately 5,600 TPS, with projections above 24,000 TPS as BPO capacity scales (Datawallet, April 2026).
Stage 2: PeerDAS (Fusaka upgrade, December 2025)
PeerDAS (EIP-7594) introduced Data Availability Sampling to Ethereum's mainnet for the first time. Instead of requiring every node to download complete blobs, nodes sample small random subsets. Combined with erasure coding, which adds redundancy so data can be reconstructed even if portions are missing. This allows Ethereum to safely increase blob count without proportionally increasing node bandwidth requirements.
The practical result: blob capacity can scale further without centralizing the validator set.
Stage 3: Full Danksharding (Future)
The long-term target is ~128 blobs per slot, representing approximately 16 MB of DA capacity per block, roughly a 20x increase over current post-BPO capacity. Full Danksharding is not traditional sharding (no splitting of execution); it distributes only DA data across nodes, keeping the architecture simpler and more secure than earlier sharding proposals.
Prerequisites include PeerDAS maturing at scale and Proposer-Builder Separation (PBS) becoming fully enshrined in the protocol.
Ethereum DA Roadmap at a Glance
Stage | Upgrade | Blob target/block | Approx. DA capacity |
| Current | EIP-4844 + BPO forks | 14 target / 21 max | ~1.8 MB/slot |
| Next | Full PeerDAS scaling | Significant increase | Multi-MB/slot |
| Future | Full Danksharding | ~128 | ~16 MB/slot |
What this means for builders:
Rollups choosing alternative DA layers today are primarily choosing cost. If Ethereum's native DA capacity scales as planned, that cost gap narrows substantially. The question becomes: how much are you paying in trust assumptions for the savings you are capturing today?
- For rollups building toward long-term decentralization and security guarantees, staying on Ethereum native DA and waiting for the roadmap to deliver may be the lower-risk path.
- For rollups that need low fees now and can accept additional trust trade-offs, Celestia or EigenDA are legitimate choices with real traction.
The DA market in 2026 is competitive precisely because Ethereum's roadmap is not yet fully delivered. That gap is the window that alternative DA layers are racing to fill.
ByteByByte's Perspective
Ethereum is deliberately stepping back from execution. The rollup-centric roadmap is a deliberate architectural choice to become the trust anchor for an entire ecosystem of chains rather than competing to be the fastest execution environment. Every improvement to blob capacity is Ethereum doubling down on being the most reliable place to prove that data existed and was available at a specific point in time.
If Ethereum succeeds in making its DA capacity abundant and cheap, it does not need to execute a single transaction to remain the most important blockchain in the ecosystem. The race between Ethereum's DA roadmap and modular competitors like Celestia is really a race over whether Ethereum can deliver that abundance before rollup teams permanently route around it.
Sources and Further Reading
- Ethereum.org – "Data Availability" https://ethereum.org/developers/docs/data-availability/
- Ethereum.org – "Scaling" https://ethereum.org/roadmap/scaling/
- Ethereum.org – "Danksharding" https://ethereum.org/roadmap/danksharding/
- Ethereum.org – "PeerDAS" https://ethereum.org/roadmap/fusaka/peerdas/
- EIP-4844.com – "EIP-4844: Shard Blob Transactions" https://www.eip4844.com/
FAQs About Ethereum Data Availability for Rollups
No. If the DA layer is unavailable, users cannot access the transaction data needed to reconstruct rollup state or submit proofs. This is why DA guarantees are non-negotiable for rollup security. A rollup with unreliable DA is effectively a trusted system, not a trustless one.
