Data availability (DA) is the guarantee that a blockchain’s transaction data has been published and can be retrieved by anyone who needs it to verify the chain’s state. It sounds like a storage problem but it’s a security problem: if data is not available, no one can verify that a rollup or sidechain’s state transitions were honest, even if the proof or signatures look valid.

Why data availability is a scaling bottleneck

Ethereum’s original design puts both execution (computing state changes) and data availability (publishing transaction data) on the same chain. Every full node downloads and stores all transaction data to verify the chain. This is expensive, Ethereum’s block size is deliberately limited to keep node requirements manageable. Rollups improve execution scaling but still depend on Ethereum for data: an optimistic or ZK rollup posts compressed transaction data to Ethereum so that anyone can reconstruct the rollup’s state and verify it. Before EIP-4844, this data was posted as calldata, which is expensive, roughly $0.50–$2.00 per transaction in L2 fees during high-demand periods.

EIP-4844 (introduced in March 2024) added “blobs”, a new, cheaper data format specifically for rollup data. Blobs are not stored permanently by Ethereum nodes (they are pruned after about 18 days) but are available long enough for fraud proofs to be submitted or ZK proofs to be verified. The result: L2 transaction fees dropped 80–95% immediately after EIP-4844 activation. Arbitrum fees went from $0.08–$0.30 to $0.001–$0.01 per transaction. The full mechanism of EIP-4844 is in our EIP-4844 explainer.

What this means for traders

DA layers are a new infrastructure category betting that Ethereum blobs will eventually become the bottleneck again at scale. Celestia launched in 2023 as a purpose-built DA layer: it does nothing but order and publish data cheaply, delegating execution to rollups that use it as their DA layer instead of Ethereum. EigenDA (built on EigenLayer restaking) and Avail are competitors. Rollups using alternative DA layers (like those in the “rollup as a service” ecosystem on Celestia) achieve even lower fees than Ethereum-native rollups but make a security tradeoff: their DA guarantee depends on Celestia’s validator set rather than Ethereum’s, which is smaller and less battle-tested.

For traders, the practical implication is fee sensitivity: rollups on Ethereum DA (Arbitrum, Base, Optimism) have predictable costs tied to Ethereum blob demand. Rollups on Celestia or EigenDA have costs tied to those smaller networks’ demand, which is currently very low but theoretically could spike. Security-conscious users prefer Ethereum DA for high-value transactions; cost-sensitive applications (gaming, micro-transactions) increasingly use alternative DA. Understanding how L2 rollups sit on top of this DA infrastructure connects to our Layer 2 explainer and our ZK rollups guide.

A concrete example

A gaming protocol wants to handle 10,000 transactions per second for in-game asset trades. On Ethereum mainnet: impossible (15 TPS maximum). On an optimistic rollup using Ethereum DA (Arbitrum): feasible at low fees post-EIP-4844, but at 10,000 TPS the blob demand would saturate Ethereum’s blob capacity and push fees back up significantly. On a Celestia-based rollup (like Eclipse): the same 10,000 TPS at a fraction of the cost, with the security assumption that Celestia’s validators remain honest and available. The Celestia network has about $1B in staked value securing it vs Ethereum’s $50B+, a 50x security difference. For $0.01 NFT trades, this tradeoff is acceptable. For $1M asset settlements, it probably isn’t.

Frequently asked questions

What is a data availability attack? A DA attack is when block producers withhold transaction data: they produce a valid-looking block but don’t release the data needed to verify it. Light clients who only check block headers (not full data) cannot detect this. DAS (data availability sampling) solves it: light clients can probabilistically verify full availability by sampling random small chunks of each block, requiring 99.9%+ of data to be available without downloading everything.

Is Ethereum DA always better than alternatives? For high-value use cases, yes, Ethereum has significantly more economic security and a longer track record. For high-volume, low-value applications where fees matter most and values at risk per transaction are small, alternative DA layers offer a rational cost-security tradeoff.

What happens to a rollup if its DA layer goes down? If the DA layer fails to publish data (network outage, censorship, or attack), the rollup cannot finalize new transactions on its settlement layer. Pending transactions are stuck until DA is restored. This is a liveness failure, not necessarily a security failure, funds already on-chain are not lost, just temporarily inaccessible. Settlement-layer security depends on whether the DA layer can be verified to have been available, not whether it is currently up.