Data availability (DA) answers a deceptively simple question: can anyone verify that all transaction data for a blockchain has been published? This matters because rollups — the primary scaling solution for Ethereum — process transactions off-chain but must post the data somewhere publicly accessible. Without verifiable data availability, a rollup operator could process invalid transactions and no one would know, since they couldn't access the data needed to generate fraud proofs or verify zero-knowledge proofs. DA has become the key bottleneck for rollup scalability, and dedicated DA layers like Celestia are among the most impactful infrastructure projects in crypto.
When rollups post data to Ethereum, they use calldata or blobs (post-EIP-4844), which compete for limited blockspace. Even with blobs reducing costs significantly compared to calldata, Ethereum can only handle about 750 KB per block of blob data — roughly 6 blobs of 128 KB each. As rollup adoption grows, demand for blob space increases, driving up costs. At peak usage, blob fees have spiked to make rollup transactions significantly more expensive. This ceiling on Ethereum's DA capacity means either rollups accept higher costs, or they look for alternative DA layers that can offer more space at lower prices.
Celestia is the pioneer of modular DA, offering a blockchain optimized solely for data availability. It uses data availability sampling (DAS) — light nodes randomly sample small chunks of block data to probabilistically verify that the full data was published, without downloading everything. This allows Celestia to scale DA capacity by adding more light nodes. EigenDA leverages Ethereum's existing validator set through EigenLayer restaking, offering DA secured by restaked ETH — a different trust model than Celestia's independent validator set. Avail, built by ex-Polygon founders, provides DA with a focus on validity proofs and light client verification. Each solution offers different security assumptions, costs, and performance characteristics.
Cheap DA has profound implications. When posting data costs fractions of a cent instead of dollars, rollups can offer near-zero transaction fees, making on-chain applications competitive with centralized alternatives. This enables use cases that were economically impractical before — on-chain gaming, social media, micropayments, and IoT data streams. The tradeoff is security: data posted to Celestia or EigenDA doesn't have the same guarantees as data posted to Ethereum mainnet. Each DA layer has its own trust model, validator set, and potential failure modes. For high-value financial applications, Ethereum DA may remain preferred despite the cost. For consumer applications where cost matters more than maximum security, alternative DA layers unlock a new design space.
Data availability (DA) ensures that transaction data is published and accessible to anyone who wants to verify the chain or rebuild state. This sounds basic but is fundamental to scaling. If a rollup withholds transaction data, users cannot prove their balances or exit the rollup if it goes offline — funds become stuck. Data availability ensures users can always verify their state and recover funds in worst-case scenarios. For Ethereum L2s, DA cost is the dominant fee component. Cheap DA means cheap L2 transactions; expensive DA means expensive L2 transactions. The EIP-4844 (proto-danksharding) upgrade reduced DA costs by 10-100x, dropping L2 fees dramatically and making Ethereum genuinely cheap for everyday use.
Rollups can choose where to post data with different cost/security trade-offs. Ethereum mainnet (calldata) is the most secure but most expensive. Ethereum blobs (post-EIP-4844) are much cheaper while maintaining Ethereum's security. Celestia is a dedicated DA layer with 100x+ cost reduction versus Ethereum, used by chains like Manta, Movement, and others. EigenDA leverages restaked ETH for DA security, integrated with Ethereum's economic security. Avail (formerly Polygon Avail) provides similar dedicated DA. NearDA leverages NEAR's existing infrastructure. Each represents different security/cost trade-offs. The trend: rollups using non-Ethereum DA are technically 'validiums' or 'optimums' rather than pure rollups, with weaker security guarantees but much lower costs.
Data Availability Sampling (DAS) is a technique that lets light clients verify data is available without downloading all of it. Each light client randomly samples small pieces; if data is available, all samples succeed; if data is withheld, sampling reveals it probabilistically. With enough light clients sampling, withholding data becomes detectable even for massive datasets. This is fundamental to Ethereum's danksharding roadmap — full danksharding will use DAS to scale DA to handle hundreds of thousands of transactions per second across rollups while maintaining security. Celestia already uses DAS in production. The technique enables scalability that would otherwise require full nodes to download enormous data volumes, threatening decentralization.
Cost. Even after EIP-4844, Ethereum DA is significantly more expensive than alternatives like Celestia. For high-throughput applications, this difference matters — gaming, social, and consumer apps need very cheap transactions. The trade-off is security: Ethereum DA inherits Ethereum's security; alternative DA layers have their own trust assumptions. Different applications make different trade-offs.
They have weaker security guarantees than rollups using Ethereum DA. If the alternative DA layer fails or goes offline, validium users could lose ability to verify their state. The risk is real but bounded — major DA layers have meaningful security infrastructure, and emergency exits to Ethereum are still possible. For most applications, the cost savings outweigh the marginal security reduction. For highest-value applications, Ethereum DA remains the choice.
Danksharding is Ethereum's planned architecture for massive DA scaling. Proto-danksharding (EIP-4844) introduced blobs — temporary data that's much cheaper than calldata. Full danksharding will scale this to ~125x current capacity through DAS, allowing rollups to collectively process hundreds of thousands of TPS while maintaining Ethereum's security. The roadmap spans multiple years; significant scaling improvements have already arrived.