Data Availability & Modular Blockchains Guide 2026
Data Availability (DA) layers are the cornerstone of blockchain scalability in 2026. As rollups scale to handle thousands of transactions per second, they generate terabytes of transaction data annually. Dedicated DA layers like Celestia, EigenDA, and Avail solve the storage and verification problem, enabling a modular blockchain architecture where execution, consensus, settlement, and data availability are separate layers. Celestia commands 50% of the DA market with 160GB+ processed, EigenDA achieves 100MB/s throughput via Ethereum restaking, and Avail secures multichain integrations. This guide explores the economics, technology, and competitive landscape of Data Availability in 2026.
1. What Is Data Availability?
Data Availability (DA) is the guarantee that all transaction data in a block is publicly accessible, verifiable, and retrievable by network participants. In monolithic blockchains like Ethereum, data availability is handled by the same consensus layer that orders transactions: validators store full blocks, and light clients can download block data to verify transactions. In modular blockchains, DA becomes a separate, specialized layer.
We wrote this guide because the existing explanations online are either too simplified or assume PhD-level knowledge. Neither serves most readers.
For rollups (Arbitrum, Optimism, Starkware), DA is critical infrastructure. Rollups execute thousands of transactions off-chain, compress them, and post transaction data to an on-chain settlement layer. The settlement layer relies on DA to verify that rollup operators didn't hide or censor transactions. If transaction data becomes unavailable, users cannot prove rollup state is correct, creating a trust failure. DA layers solve this: they store rollup data cheaply and efficiently, with cryptographic proofs that data is available and can be reconstructed.
Ethereum processes ~15 transactions per second (TPS) on mainnet due to data constraints—blocks are limited to ~125KB. Rollups batch transactions and post only compressed data to Ethereum, achieving 100-500 TPS per rollup. However, as rollups scale, they generate 10-100GB of transaction data monthly. Ethereum cannot store this volume economically (cost would exceed $100M annually). DA layers like Celestia store 160GB+ of rollup data at 1/100th the cost. The result: rollups achieve 1000s TPS at minimal cost, unlocking Ethereum scalability. Without DA layers, rollups would remain bottlenecked by Ethereum settlement throughput.
The core innovation of 2026 DA layers is efficiency. Instead of requiring nodes to download entire blocks (100MB), light clients use Data Availability Sampling (DAS) to randomly sample small chunks across the block. Cryptographic proofs (erasure coding) guarantee that if enough samples can be collected, the original data can be reconstructed. This reduces bandwidth requirements from 100MB to 1-2MB per block for light clients—a 50-100x improvement.
2. The Modular Blockchain Thesis
The modular blockchain thesis is foundational to understanding why DA layers exist. Monolithic blockchains (Bitcoin, Ethereum, Solana) conflate four distinct functions into a single protocol layer: execution (processing transactions), consensus (reaching agreement on transaction order), settlement (finalizing state and handling disputes), and data availability (storing and verifying transaction data). This all-in-one design is secure but inefficient—each layer is limited by the slowest bottleneck.
Modular blockchains decouple these layers. Rollups handle execution independently, processing transactions without consensus. A separate DA layer stores transaction data and proves availability. A settlement layer (Ethereum, Solana, Cosmos Hub) finalizes state. A consensus layer orders data on the DA chain. This separation enables radical optimization: rollups achieve 1000s TPS because they avoid consensus delays. DA layers specialize in cheap storage and verification. Settlement layers focus on security and finality.
The Four Layers
Execution Layer: Rollups (Arbitrum, Optimism, StarkNet) batch and compress user transactions. No consensus is needed—rollups simply post batches to the DA layer. Speed: 100-4000 TPS per rollup. Cost: $0.01-0.10 per transaction.
Data Availability Layer: Celestia, EigenDA, Avail store rollup transaction data and prove it's accessible via DAS. Security is high—DA layers use strong consensus (Proof-of-Stake, Bitcoin consensus) to prevent data withholding attacks. Cost: $0.0001-0.001 per KB of data.
Settlement Layer: Ethereum, Solana, or other chains finalize rollup state. Settlement performs fraud proofs (optimistic rollups) or validity proofs (zk-rollups) to confirm rollup execution is correct. Settlement is expensive—only rollup commitments, not full transactions, are settled. Cost: ~$1-100 per rollup update.
Consensus Layer: DA layers (Celestia, Avail) operate their own consensus to order and finalize data blocks. EigenDA uses Ethereum consensus via restaking. Consensus ensures data is available in the correct order, preventing reorgs that could break rollup finality.
Ethereum mainnet: ~15 TPS, ~125KB blocks, ~100 transactions per block. Single Arbitrum rollup: ~250 TPS using Celestia DA, ~16MB batches, ~2000 transactions per batch. With 10 rollups on Celestia: 2500 TPS collective throughput at 1/10th the Ethereum L1 cost. With 100 rollups: 25,000 TPS. Modular architecture enables near-linear scalability—each new rollup adds throughput without congesting shared resources.
3. How Data Availability Sampling Works
Data Availability Sampling (DAS) is a cryptographic breakthrough enabling light clients to verify data availability without downloading entire blocks. The key insight: if a light client randomly samples enough chunks from a block and can reconstruct it via erasure coding, the data is available—even if the client never sees 100% of the data. An attacker hiding 51% of data will almost certainly be caught by random sampling.
Erasure Coding Primer
Erasure codes (like Reed-Solomon) transform original data into redundant coded pieces such that any 50% of the pieces can reconstruct the original. Example: if you have 100 pieces of data, an erasure code produces 200 coded pieces. Any 100 of those 200 pieces can reconstruct the original data. An attacker hiding 101 pieces (51%) will make reconstruction impossible. A light client sampling 30 random pieces has a 99.9% probability of collecting 15+ non-hidden pieces, proving data is available.
2D Commitment Schemes
Celestia and Avail use 2D Reed-Solomon: data is arranged in a matrix (e.g., 128 rows × 128 columns), with erasure coding applied both horizontally and vertically. Light clients sample along rows and columns. This design allows efficient fraud proofs: if a block producer withholds data in a region, honest nodes can download that region, compute cryptographic proofs of withholding, and broadcast them to prove the attack. This incentivizes light clients to sample aggressively.
Sample Count: Celestia light clients typically perform 16-32 samples per block. Each sample is ~2-5KB. Total bandwidth: 32-160KB per block. Full block size: 128MB. Savings: 99.9%. With this efficiency, a smartphone can sync and verify a DA layer chain in real time.
DAS assumes a "1/3 honest majority" of light clients. If at least 1/3 of light clients are honest and online, an attacker cannot hide data from all of them—at least one honest client will sample a missing piece and raise an alert. This assumption is realistic for global networks with thousands of light clients. DAS is not suitable for closed, small-validator networks. DAS also assumes block producers cannot commit to data they haven't produced—light clients fetch data from network peers, and if peers are unavailable, data is truly unavailable.
4. Celestia Deep Dive
Celestia is the dominant DA layer in 2026, commanding approximately 50% of the market. Founded in 2020, Celestia launched mainnet in October 2023 and has grown rapidly. The chain uses Bitcoin as its consensus layer (through a light client embedded in Celestia validator software), ensuring independent security. As of April 2026, Celestia has processed 160GB+ of rollup data from Arbitrum, Optimism, Polygon, and emerging rollups.
The Matcha Upgrade
The Matcha upgrade, deployed in Q1 2026, doubled Celestia's block size from 64MB to 128MB, effectively doubling throughput. Block time remains 6 seconds, so network capacity increased to ~21.3 MB/s or ~160 Gbps worth of data (theoretical maximum). Matcha also optimized the 2D Reed-Solomon commitment scheme for larger block sizes, maintaining DAS security with 99.9% bandwidth savings. Additionally, Matcha reduced inflation from 5% to 2.5% annually by slashing validator rewards—TIA incentive alignment shifted toward application fees rather than staking rewards.
Fibre Blockspace Protocol
Fibre is Celestia's response to throughput demands. The Fibre protocol promises 1 terabit/s throughput, a 1,500x improvement over previous roadmap targets. This is achieved via optimized block propagation and pipelining (overlapping block production and consensus rounds). If realized, Fibre enables Celestia to support 1,000+ concurrent rollups each processing 1000+ TPS—supporting an entire Ethereum-scale ecosystem on a single DA layer.
Tokenomics & Market Metrics
TIA token is the network's utility and staking token. Current price: ~$0.30 (April 2026). Market cap: ~$262M. Supply: ~873M tokens fully diluted. Staking yield: ~8-12% annually. Transaction fees are denominated in TIA and burned, reducing supply over time. As rollup adoption grows, demand for blockspace (and thus TIA fees) is expected to increase substantially. Celestia's 50% market dominance and proven infrastructure make TIA an essential utility token for rollup deployments.
Early market entry: Launched mainnet in October 2023, 1-2 years ahead of competitors. Bitcoin consensus: Uses Bitcoin's security and decentralization, appealing to crypto natives. Proven reliability: Processed 160GB+ with zero downtime or finality reversions. Developer ecosystem: Strong tooling, documentation, and community. Token distribution: Fair initial distribution and progressive unlock schedule. Upgrade cadence: Regular upgrades (Matcha) showing commitment to performance. Risk: Extreme inflation (5% → 2.5%) and rising competition from EigenDA and Avail.
5. EigenDA Deep Dive
EigenDA is a Data Availability layer built on EigenLayer, Ethereum's restaking protocol. While Celestia operates independent consensus, EigenDA leverages Ethereum's existing validator set. Validators stake ETH in EigenLayer and commit to running EigenDA nodes, storing data, and providing attestations. This design provides Ethereum-native security: EigenDA inherits the security of Ethereum's 40M+ ETH staked, worth ~$100B+.
DAC Model & Throughput
EigenDA uses a Data Availability Committee (DAC) model: 32 EigenLayer operators (the most-staked validators) attest to data availability. Light clients only need to download 32 signatures (~1KB) instead of full data blocks (100MB). This is highly efficient—bandwidth savings are even greater than DAS (99.99% vs. 99.9%). Throughput is currently 100MB/s, comparable to Celestia's Matcha (128MB × 10 sec blocks ≈ 12.8MB/s post-Matcha, but with larger block sizes Celestia could reach 100MB/s too). EigenDA is optimized for high-throughput, trusted validators rather than light client verification.
Ethereum-Native Integration
EigenDA's core advantage is Ethereum alignment. Rollups settle on Ethereum and use EigenDA for DA—the entire stack stays on Ethereum. This simplifies security analysis: if EigenDA goes down, settlements are delayed but not reversed (Ethereum settlement is independent). If an EigenDA operator becomes malicious, slashing is enforced by Ethereum—the operator loses staked ETH. This tight integration appeals to Ethereum ecosystem rollups (Arbitrum, Optimism, zkSync).
EigenDA's pricing model differs from Celestia: instead of per-byte fees, users pay for the time data is stored (e.g., 7 days of DA). This creates stable, predictable costs for rollups rather than the variable pricing of blockspace-based systems.
Celestia: Independent consensus, proven infrastructure, Bitcoin-backed security. EigenDA: Ethereum-native, restaking-based, high throughput, integrated settlement. Celestia excels for multichain rollups. EigenDA excels for Ethereum L2s. Neither has a clear winner—market will support both. Celestia's advantage: independence and ecosystem maturity. EigenDA's advantage: Ethereum alignment and validator economics.
6. Avail Deep Dive
Avail is a Data Availability layer focused on multichain scalability. Unlike Celestia (Bitcoin consensus) and EigenDA (Ethereum-based), Avail operates an independent consensus but positions itself as agnostic infrastructure. Avail's key differentiator is multichain integration: rollups can publish DA to Avail while settling on their preferred chain (Arbitrum, Optimism, Polygon, or any other).
Multichain Settlement Strategy
Avail secured integrations with Arbitrum, Optimism, and Polygon, the three largest rollup ecosystems. This enables rollups to reduce settlement latency: instead of waiting for Ethereum L1 finality (~15 minutes), rollups can finalize on Avail (~6 seconds) while maintaining economic security. Rollups also reduce costs by using Avail's cheaper DA rather than Ethereum mainnet. This multichain approach appeals to rollups wanting independence from Ethereum settlement while maintaining strong security.
Technical Architecture
Avail uses a similar 2D Reed-Solomon scheme to Celestia, enabling efficient DAS. Block time is 6 seconds, and current block size is 128MB, matching Celestia post-Matcha. Throughput: ~21.3 MB/s or ~160 Gbps theoretical maximum. Avail's consensus layer is Proof-of-Stake with ~250 validators (smaller than Celestia but still decentralized). Security is strong but more specialized than Celestia's Bitcoin-backed design or EigenDA's Ethereum-backed design.
Avail's pricing is also blockspace-based (per byte), similar to Celestia. Costs are competitive with Celestia due to lower network congestion and smaller validator set. For multichain rollups, Avail is cost-efficient.
Avail targets the "middle market" of rollups—those wanting DA benefits without committing to Ethereum settlement or Celestia's independent consensus. Avail is well-suited for rollups planning multichain deployments or preferring settlement on Polygon/Arbitrum. Growth depends on rollup adoption and continued investment in Avail validator infrastructure. Risk: Avail is newer than Celestia and smaller than EigenDA in terms of security backing.
7. DA Layer Comparison
The DA layer landscape is diverse, with three major players and unique tradeoffs in security, throughput, and integration.
| Aspect | Celestia | EigenDA | Avail |
|---|---|---|---|
| Consensus Layer | Bitcoin (light client embedded) | Ethereum (via EigenLayer) | Independent PoS (250 validators) |
| Security Model | Proof-of-Stake on Bitcoin consensus | DAC (32 top validators) | Proof-of-Stake |
| Throughput | ~21 MB/s (Matcha upgrade) | ~100 MB/s (DAC model) | ~21 MB/s |
| Block Time | 6 seconds | 12 seconds (Ethereum finality) | 6 seconds |
| DA Cost | Blockspace-based (~$0.0001/KB) | Time-based (7-day storage contract) | Blockspace-based (~$0.0001/KB) |
| Light Client Bandwidth | ~100KB per block (DAS) | ~1KB per block (DAC) | ~100KB per block (DAS) |
| Settlement Integration | Agnostic (rollups settle independently) | Ethereum-native | Multichain (Ethereum, Arbitrum, Polygon) |
| Market Share (April 2026) | ~50% (Arbitrum, Optimism, Polygon) | ~30% (Ethereum L2 preference) | ~15% (emerging multichain rollups) |
| Token Price / Market Cap | TIA ~$0.30 / $262M | No native token (restaking rewards) | AVL ~$0.15 / ~$50M (estimate) |
Celestia: Cost efficiency and proven infrastructure. Best for cost-optimized rollups and multichain ecosystems. EigenDA: Ethereum alignment and maximum security. Best for Ethereum L2s and applications requiring tight protocol integration. Avail: Multichain flexibility and settlement optionality. Best for rollups wanting DA benefits without committing to a single settlement chain. No single winner—the market will support all three based on rollup preferences.
8. Future of Data Availability
The DA layer market is nascent. As of April 2026, adoption is strong but growth is just beginning. The future of DA hinges on several factors: (1) rollup demand growth, (2) throughput improvements, (3) cost reduction, and (4) security evolution.
Throughput Roadmap
Celestia's Fibre protocol promises 1 terabit/s throughput by 2027. If realized, this represents a 50x improvement over current ~21 MB/s. EigenDA is pursuing higher throughput via improved DAC designs and operator scaling. Avail is exploring sharding strategies to partition data across multiple consensus layers. By 2027, all three should support 1-2 GB/s throughput, enabling 10,000+ concurrent rollups each processing 1000+ TPS.
Cost Trajectory
Current DA costs for rollups are ~$0.0001-0.001 per byte. Historical Ethereum mainnet costs are ~$10-100 per kilobyte. DA layers represent a 100-1000x cost reduction. As DA layer throughput scales, costs should drop further—Moore's law applies to blockchain data infrastructure. By 2027, expect DA costs to fall to $0.00001/byte or lower, making DA almost free relative to execution and settlement costs.
Interoperability & Bridging
Future DA layers may support interoperability: rollups could use data from multiple DA layers, mixing Celestia and EigenDA or Avail in a single transaction. Cross-layer bridges would verify DA proofs, enabling flexible deployments. This could lead to a "DA neutral" ecosystem where rollups choose DA layers dynamically based on cost and latency.
Security Evolution
As DA layers mature, security assumptions will be tested. Celestia's Bitcoin-backed consensus has been robust, but larger data loads may reveal edge cases. EigenDA's DAC model works well for high-throughput but is more centralized than full DAS. Avail's independent consensus is still unproven at scale. By 2027-2028, the market will have clarity on which security models are most resilient. The winner may depend on unforeseen attacks or regulatory actions affecting validator incentives.
The DA market will likely consolidate to 2-3 major players by 2027. Celestia's early lead and Bitcoin consensus are strong advantages. EigenDA's Ethereum backing is powerful. Avail's multichain positioning is differentiated. The wildcard: regulatory changes affecting staking (especially EigenDA), or technical breakthroughs in DAS or DAC designs. The ultimate winner will be the DA layer achieving the best tradeoff of security, throughput, and cost at scale.
9. Frequently Asked Questions
Can rollups use multiple DA layers simultaneously?
In theory, yes. A rollup could post data to Celestia and EigenDA in parallel, achieving redundancy and load balancing. In practice, this adds complexity and cost. Most rollups use a single primary DA layer and a secondary fallback. As cross-layer bridges mature, hybrid strategies may become common. For now, rollups choose one DA layer based on security, cost, and integration preferences.
What happens if a DA layer goes down?
If a DA layer stops producing blocks, rollups cannot post new transaction data. New transactions cannot be executed, but existing state remains finalized. Rollups may exit to a fallback settlement layer (Ethereum mainnet) to unwind positions. A DA layer outage is serious but not a complete failure if fallbacks exist. Redundancy through multiple DA layers reduces this risk. To date, Celestia, EigenDA, and Avail have not experienced extended outages.
How do light clients on DA layers verify transactions?
Light clients use Data Availability Sampling: they randomly request small chunks (~2-5KB) of block data from the network and verify erasure-coded commitments. If enough samples are collected without errors, the light client is confident the entire block is available. Light clients don't verify transaction semantics—that's the settlement layer's job. DA layer light clients only verify data availability and block ordering.
Why is modular architecture better than monolithic blockchains?
Modular architecture allows specialization: execution layers (rollups) focus on transaction processing, DA layers focus on cheap storage, settlement layers focus on security. Each layer can optimize independently. Monolithic blockchains must make universal tradeoffs (e.g., Ethereum chose 12-second blocks to balance security and throughput). Modular systems achieve better scalability—Ethereum mainnet processes ~15 TPS, but a rollup using modular DA can process 100-1000 TPS. The tradeoff: modularity adds complexity and requires coordination between layers.
Which DA layer should I use for my rollup?
Choose based on your priorities: For cost efficiency and multichain support: Celestia. For Ethereum alignment and maximum security: EigenDA. For multichain settlement optionality: Avail. If launching on Ethereum L2: EigenDA is recommended due to tight integration. If building a multichain rollup: Celestia or Avail are better. No single "best" choice—evaluate your application's security, cost, and integration requirements.
Can DA layers be hacked or attacked?
DA layers can experience attacks like any blockchain: 51% consensus attacks, DAS vulnerabilities, or sequencer failures. Celestia's Bitcoin-backed consensus is very hard to attack—you'd need 51% of Bitcoin hash power. EigenDA's DAC model is more vulnerable if 17+ of 32 validators collude. Avail's independent consensus is vulnerable if validators collude. However, DA attacks don't steal funds directly—they delay data availability, which is detected and can trigger fallback settlement. The best defense: use a DA layer with strong consensus (Celestia, EigenDA on Ethereum) and redundant settlement options.
Further Learning
Deepen your understanding of blockchain infrastructure and related topics:
Educational disclaimer: This guide is for informational purposes only and does not constitute financial advice. Crypto involves significant risk — do your own research before making any decisions. Learn more about our team.
Educational disclaimer: This guide is for informational purposes only and does not constitute financial advice. Crypto involves significant risk — do your own research before making any decisions. Learn more about our team.