Data availability is ~95% of the costs that rollups pay to post data on Ethereum. It's the primary scaling bottleneck and the reason modular blockchains exist. Understanding DA is essential for evaluating rollups, alt-DA solutions, and the modular vs. monolithic debate.
What is Data Availability?
Data availability (DA) is the guarantee that the data needed to verify a block was actually published and is accessible to anyone who needs it. In simpler terms: did the block producer actually make the transaction data available, or did they hide it?
This might seem trivial, but it's a fundamental security requirement:
- Without DA: A malicious block producer could commit to a state transition without publishing the underlying data. No one could verify if the transition was valid.
- With DA: Anyone can download the data, re-execute transactions, and verify the state transition is correct.
Why DA Matters for Rollups
Rollups execute transactions off-chain but must post data somewhere so the main chain can verify correctness. This creates a direct relationship between DA costs and rollup transaction fees.
The Cost Breakdown
When a rollup posts data to Ethereum:
- Data posting: ~95% of rollup operating costs
- State updates: ~4% of costs
- Verification: ~1% of costs
Real cost comparison (2025 data):
| DA Layer | Cost per MB | Relative Cost |
|---|---|---|
| Ethereum calldata | ~$50-100 | Baseline |
| Ethereum blobs (EIP-4844) | ~$3.83 | ~25x cheaper |
| Celestia | ~$0.07 | ~55x cheaper than blobs |
| EigenDA | ~$0.01-0.05 | ~75x cheaper than blobs |
Eclipse (a rollup using Celestia for DA) reported data posting costs of $0.07/MB compared to $3.83/MB for Ethereum blobs—a 55x reduction. This translates directly to cheaper user transactions.
Data Availability Sampling (DAS)
The breakthrough that enables scalable DA is Data Availability Sampling (DAS). Instead of downloading all data, nodes randomly sample small portions and use cryptographic proofs to verify the whole block's availability.
How DAS Works
- Erasure coding: Block data is encoded with redundancy (like RAID for your hard drive). Even if 50% of the data is missing, it can be reconstructed.
- Random sampling: Light nodes request random small chunks of the block
- Probability math: If a light node successfully retrieves several random samples, there's overwhelming probability the full data is available
- Network effect: More light nodes sampling = higher confidence = safer to increase block size
The Math Behind DAS
Suppose a block has 100 chunks. A malicious producer wants to hide 1 chunk (1% of data). If a light node randomly samples 30 chunks:
- Probability of missing the hidden chunk: 0.99^30 ≈ 74%
- But with 1000 light nodes each sampling 30 chunks, probability that ALL miss it: 0.74^1000 ≈ 0%
The more light nodes participate, the harder it is to hide any data.
Monolithic vs. Modular Blockchains
The Monolithic Approach
Traditional blockchains (Bitcoin, Ethereum pre-Dencun) handle everything on one layer:
- Execution: Running transactions
- Settlement: Finalizing state
- Consensus: Agreeing on block order
- Data Availability: Ensuring data is published
Problem: All functions compete for the same limited blockspace. Can't optimize one without affecting others.
The Modular Approach
Modular blockchains separate these functions into specialized layers:
| Layer | Function | Examples |
|---|---|---|
| Execution | Run transactions, compute state | Arbitrum, Optimism, zkSync |
| Settlement | Verify proofs, finalize state | Ethereum L1 |
| Consensus | Order transactions | Shared sequencers |
| Data Availability | Ensure data is published | Celestia, EigenDA, Avail |
Benefit: Each layer can scale independently. A DA layer can handle massive throughput without affecting execution costs.
DA Layer Comparison
Celestia
The first production DA layer, launched October 2023. Celestia is purpose-built for data availability using DAS.
Key features:
- DAS-native: First chain built around data availability sampling
- Namespaced Merkle Trees: Different rollups get their own data "namespaces"
- Light node verification: Phones can verify DA without downloading full blocks
- Sovereignty: Rollups maintain full control, no Celestia-imposed rules
Current specs:
- Block time: ~12 seconds
- Data throughput: ~2-6 MB per block
- Cost: ~$0.07/MB
- Finality: ~10 minutes (CometBFT)
EigenDA
EigenDA leverages Ethereum's restaking infrastructure (EigenLayer) to provide DA with Ethereum-aligned security.
Key features:
- Restaking security: ETH validators restake to secure DA
- Ethereum alignment: Inherits economic security from Ethereum
- No DAS: Uses different verification model based on restaked collateral
- High throughput: Designed for hyperscale data
Current specs:
- Throughput: 10+ MB/s target
- Cost: ~$0.01-0.05/MB
- Security: Backed by restaked ETH
Avail
Another DAS-based DA layer with focus on validity proofs and interoperability.
Key features:
- KZG commitments for efficient proofs
- Focus on ZK rollup support
- Bridge to Ethereum for settlement
Ethereum Blobs (EIP-4844)
Ethereum's native DA solution, implemented in the Dencun upgrade (March 2024). Blobs are a new transaction type that stores temporary data.
Key features:
- Native Ethereum security: Same validators secure blobs
- Temporary storage: Data pruned after ~18 days
- Limited throughput: ~125 KB per block currently
- Future scaling: Full danksharding will add DAS
Comparison Table
| Property | Celestia | EigenDA | Ethereum Blobs |
|---|---|---|---|
| DAS Support | Yes (core feature) | No | Future (danksharding) |
| Throughput | 2-6 MB/block | 10+ MB/s | ~125 KB/block |
| Cost | ~$0.07/MB | ~$0.01-0.05/MB | ~$3.83/MB |
| Security Model | Own validator set + DAS | Restaked ETH | Ethereum consensus |
| Finality | ~10 minutes | Ethereum-dependent | ~13 minutes |
| Best For | Sovereign rollups, modular stacks | Ethereum-aligned rollups | Ethereum-native rollups |
Security Considerations
DA vs. Data Storage
Important distinction: DA ≠ permanent storage.
- DA: Proving data was published and available for a period
- Storage: Keeping data permanently accessible
DA layers typically don't store data forever. Celestia prunes after ~30 days. Ethereum blobs prune after ~18 days. If you need historical data, you need separate archival solutions.
Trust Assumptions
| DA Solution | Trust Assumption |
|---|---|
| Ethereum blobs | Ethereum consensus (most battle-tested) |
| Celestia | Celestia validators + DAS sampling |
| EigenDA | EigenLayer restakers (Ethereum-aligned) |
| DA Committee | Small trusted group (weakest) |
Using Celestia or EigenDA instead of Ethereum blobs is cheaper but introduces new trust assumptions. The rollup's security now depends on both Ethereum AND the DA layer. For rollups holding billions in TVL, this tradeoff requires careful consideration.
Data Withholding Attacks
A malicious DA layer could withhold data, preventing users from exiting a rollup or challenging invalid state transitions.
Mitigations:
- DAS makes withholding detectable (Celestia)
- Economic penalties for withholding (EigenDA slashing)
- Multiple DA sources as fallback
Choosing a DA Solution
Use Ethereum blobs if: Security is paramount, TVL is high, cost is secondary.
Use Celestia if: Sovereignty matters, want modular architecture, cost-sensitive.
Use EigenDA if: Want Ethereum alignment with lower costs, building Ethereum-centric rollup.
| Consideration | Ethereum Blobs | Alt-DA (Celestia/EigenDA) |
|---|---|---|
| Security | Highest (Ethereum consensus) | New trust assumptions |
| Cost | Higher (~$3.83/MB) | Much lower (~$0.01-0.07/MB) |
| Throughput | Limited (for now) | Higher |
| Ecosystem | Ethereum-native tooling | Growing, less mature |
The Future of DA
Danksharding
Ethereum's full DA scaling solution. Will implement full DAS on Ethereum itself, dramatically increasing blob capacity while maintaining decentralization.
Expected improvements:
- ~16 MB blobs per block (vs ~125 KB today)
- DAS for light client verification
- Could make alt-DA less necessary for Ethereum rollups
DA Commoditization
As DA becomes a commodity, competition will drive costs toward marginal cost of storage + bandwidth. This benefits rollups and end users regardless of which DA layer wins.
Interoperability
Future rollups may use multiple DA layers simultaneously, posting to Ethereum for maximum security and Celestia for overflow, creating redundant and cost-optimized DA.
Data availability is the invisible infrastructure that determines rollup economics. As DA costs drop, transaction fees drop, enabling new use cases. The modular thesis bets that separating DA from execution unlocks massive scaling. Whether through Ethereum's danksharding or alt-DA layers, solving DA is crucial for blockchain's next phase of growth.