The 2026 Data Availability Race: Celestia, EigenDA, and Avail's Battle for Blockchain Scalability
Every Layer 2 you use relies on a hidden infrastructure most users never think about: data availability layers. But in 2026, this quiet battlefield has become the most critical piece of blockchain scalability, with three giants—Celestia, EigenDA, and Avail—racing to process terabits of rollup data per second. The winner doesn't just capture market share; they define which rollups survive, how much transactions cost, and whether blockchain can scale to billions of users.
The stakes couldn't be higher. Celestia commands roughly 50% of the data availability market after processing over 160 gigabytes of rollup data. Its upcoming Matcha upgrade in Q1 2026 will double block sizes to 128MB, while the experimental Fibre Blockspace protocol promises a staggering 1 terabit per second throughput—1,500 times their previous roadmap target. Meanwhile, EigenDA has achieved 100MB/s throughput using a Data Availability Committee model, and Avail has secured integrations with Arbitrum, Optimism, Polygon, StarkWare, and zkSync for its mainnet launch.
This isn't just infrastructure competition—it's a battle over the fundamental economics of Layer 2 networks. Choosing the wrong data availability layer can increase costs by 55 times, making the difference between a thriving rollup ecosystem and one strangled by data fees.
The Data Availability Bottleneck: Why This Layer Matters
To understand why data availability has become blockchain's most important battlefield, you need to grasp what rollups actually do. Layer 2 rollups like Arbitrum, Optimism, and Base execute transactions off-chain to achieve faster speeds and lower costs, then post transaction data somewhere secure so anyone can verify the chain's state. That "somewhere secure" is the data availability layer.
For years, Ethereum's mainnet served as the default DA layer. But as rollup usage exploded, Ethereum's limited block space created a bottleneck. Data availability fees spiked during periods of high demand, eating into the cost savings that made rollups attractive in the first place. The solution? Modular data availability layers purpose-built to handle massive throughput at minimal cost.
Data availability sampling (DAS) is the breakthrough technology enabling this transformation. Instead of requiring every node to download entire blocks to verify availability, DAS allows light nodes to probabilistically confirm data is available by sampling small random chunks. More light nodes sampling means the network can safely increase block sizes without sacrificing security.
Celestia pioneered this approach as the first modular data availability network, separating data ordering and availability from execution and settlement. The architecture is elegant: Celestia orders transaction data into "blobs" and guarantees their availability for a configurable period, while execution and settlement happen on layers above. This separation allows each layer to optimize for its specific function rather than compromising on all fronts like monolithic blockchains.
By mid-2025, more than 56 rollups were using Celestia, including 37 on mainnet and 19 on testnet. Eclipse alone has posted over 83 gigabytes through the network. Every major rollup framework—Arbitrum Orbit, OP Stack, Polygon CDK—now supports Celestia as a data availability option, creating switching costs and network effects that compound Celestia's early-mover advantage.
Celestia's Two-Pronged Attack: Matcha Upgrade and Fibre Blockspace
Celestia isn't resting on its market share. The project is executing a two-phase strategy to cement dominance: the near-term Matcha upgrade bringing production-ready scalability improvements, and the experimental Fibre Blockspace protocol targeting 1 terabit per second of future throughput.
Matcha Upgrade: Doubling Down on Production Scale
The Matcha upgrade (Celestia v6) is currently live on the Arabica testnet with mainnet deployment expected in Q1 2026. It represents the largest single capacity increase in Celestia's history.
Core improvements include:
-
128MB block size: CIP-38 introduces a new high-throughput block propagation mechanism, increasing maximum block size from 8MB to 128MB—a 16x jump. The data square size expands from 128 to 512, and maximum transaction size grows from 2MB to 8MB.
-
Reduced storage requirements: CIP-34 cuts Celestia's minimum data pruning window from 30 days to 7 days plus 1 hour, slashing storage costs for bridge nodes from 30TB to 7TB at projected throughput levels. For rollups running high-volume applications, this storage reduction translates directly to lower operational costs.
-
Light node optimization: CIP-35 introduces pruning for Celestia light nodes, allowing them to retain only recent headers rather than the entire chain history. Light node storage requirements drop to approximately 10GB, making it feasible to run verification nodes on consumer hardware and mobile devices.
-
Inflation cut and interoperability: Beyond scalability, Matcha cuts protocol inflation from 5% to 2.5%, potentially making TIA deflationary if network usage grows. It also removes the token filter for IBC and Hyperlane, positioning Celestia as a routing layer for any asset across multiple ecosystems.
In testing environments, Celestia achieved approximately 27 MB/s throughput with 88 MB blocks in the Mammoth Mini devnet, and 21.33 MB/s sustained throughput with 128 MB blocks in the mamo-1 testnet. These aren't theoretical maximums—they're production-proven benchmarks that rollups can rely on when architecting for scale.
Fibre Blockspace: The 1 Tb/s Future
While Matcha focuses on near-term production readiness, Fibre Blockspace represents Celestia's moonshot vision for blockchain throughput. The protocol is capable of sustaining 1 terabit per second of blockspace across 500 nodes—a throughput level 1,500 times the goal set in Celestia's previous roadmap.
The core innovation is ZODA, a new encoding protocol that Celestia claims processes data 881 times faster than KZG commitment-based alternatives used by competing DA protocols. During large-scale network tests using 498 GCP machines distributed across North America (each with 48-64 vCPUs, 90-128GB RAM, and 34-45Gbps network links), the team successfully demonstrated terabit-scale throughput.
Fibre targets power users with a minimum blob size of 256KB and maximum of 128MB, optimized for high-volume rollups and institutional applications requiring guaranteed throughput. The rollout plan is incremental: Fibre will first deploy to the Arabica testnet for developer experimentation, then graduate to mainnet with progressive throughput increases as the protocol undergoes real-world stress testing.
What does 1 Tb/s actually mean in practice? At that throughput level, Celestia could theoretically handle the data needs of thousands of high-activity rollups simultaneously, supporting everything from high-frequency trading venues to real-time gaming worlds to AI model training coordination—all without the data availability layer becoming a bottleneck.
EigenDA and Avail: Different Philosophies, Different Trade-offs
While Celestia dominates market share, EigenDA and Avail are carving out distinct positioning with alternative architectural approaches that appeal to different use cases.
EigenDA: Speed Through Restaking
EigenDA, built by the EigenLayer team, has released V2 software achieving 100MB per second throughput—significantly higher than Celestia's current mainnet performance. The protocol leverages EigenLayer's restaking infrastructure, where Ethereum validators reuse their staked ETH to secure additional services including data availability.
The key architectural difference: EigenDA operates as a Data Availability Committee (DAC) rather than a publicly verified blockchain. This design choice removes certain verification requirements that blockchain-based solutions implement, enabling DACs like EigenDA to reach higher raw throughput while introducing trust assumptions that validators in the committee will honestly attest to data availability.
For Ethereum-native projects prioritizing seamless integration with the Ethereum ecosystem and willing to accept DAC trust assumptions, EigenDA offers a compelling value proposition. The shared security model with Ethereum mainnet creates a natural alignment for rollups already relying on Ethereum for settlement. However, this same dependency becomes a limitation for projects seeking sovereignty beyond the Ethereum ecosystem or requiring the strongest possible data availability guarantees.
Avail: Multichain Flexibility
Avail launched its mainnet in 2025 with a different focus: optimizing data availability for highly scalable and customizable rollups across multiple ecosystems, not just Ethereum. The protocol combines validity proofs, data availability sampling, and erasure coding with KZG polynomial commitments to deliver what the team calls "world-class data availability guarantees."
Avail's current mainnet throughput stands at 4MB per block, with benchmarks demonstrating successful increases to 128MB per block—a 32x improvement—without sacrificing network liveness or block propagation speed. The roadmap includes progressive throughput increases as the network matures.
The project's major achievement in 2026 has been securing integration commitments from five major Layer 2 projects: Arbitrum, Optimism, Polygon, StarkWare, and zkSync. Avail claims over 70 partnerships total, spanning application-specific blockchains, DeFi protocols, and Web3 gaming chains. This ecosystem breadth positions Avail as the data availability layer for multichain infrastructure that needs to coordinate across different settlement environments.
Avail DA represents the first component of a three-part architecture. The team is developing Nexus (an interoperability layer) and Fusion (a security network layer) to create a full-stack modular infrastructure. This vertical integration strategy mirrors Celestia's vision of being more than just data availability—becoming fundamental infrastructure for the entire modular stack.
Market Position and Adoption: Who's Winning in 2026?
The data availability market in 2026 is shaping up as a "winner takes most" dynamic, with Celestia holding commanding early-stage market share but facing credible competition from EigenDA and Avail in specific niches.
Celestia's Market Dominance:
- ~50% market share in data availability services
- 160+ gigabytes of rollup data processed through the network
- 56+ rollups using the platform (37 mainnet, 19 testnet)
- Universal rollup framework support: Arbitrum Orbit, OP Stack, and Polygon CDK all integrate Celestia as a DA option
This adoption creates powerful network effects. As more rollups choose Celestia, developer tooling, documentation, and ecosystem expertise concentrate around the platform.
Switching costs increase as teams build Celestia-specific optimizations into their rollup architecture. The result is a flywheel where market share begets more market share.
EigenDA's Ethereum Alignment:
EigenDA's strength lies in its tight integration with Ethereum's restaking ecosystem. For projects already committed to Ethereum for settlement and security, adding EigenDA as a data availability layer creates a vertically integrated stack entirely within the Ethereum universe.
The 100MB/s throughput also positions EigenDA well for high-frequency applications willing to accept DAC trust assumptions in exchange for raw speed.
However, EigenDA's reliance on Ethereum validators limits its appeal for rollups seeking sovereignty or multichain flexibility. Projects building on Solana, Cosmos, or other non-EVM ecosystems have little incentive to depend on Ethereum restaking for data availability.
Avail's Multichain Play:
Avail's integrations with Arbitrum, Optimism, Polygon, StarkWare, and zkSync represent major partnership wins, but the protocol's actual mainnet usage lags behind announcements.
The 4MB per block throughput (versus Celestia's current 8MB and Matcha's upcoming 128MB) creates a performance gap that limits Avail's competitiveness for high-volume rollups.
Avail's true differentiator is multichain flexibility. As blockchain infrastructure fragments across Ethereum L2s, alternative L1s, and application-specific chains, the need for a neutral data availability layer that doesn't favor one ecosystem grows. Avail positions itself as that neutral infrastructure, with partnerships spanning multiple settlement layers and execution environments.
The Economics of DA Layer Choice:
Choosing the wrong data availability layer can increase rollup costs by 55x according to industry analysis. This cost differential stems from three factors:
- Throughput limitations creating data fee spikes during demand peaks
- Storage requirements forcing rollups to maintain expensive archive infrastructure
- Switching costs making it expensive to migrate once integrated
For gaming-focused Layer 3 rollups generating massive state updates, the choice between Celestia's low-cost modular DA (especially post-Matcha) versus more expensive alternatives can mean the difference between sustainable economics and bleeding capital on data fees. This explains why Celestia is projected to dominate gaming L3 adoption in 2026.
The Path Forward: Implications for Rollup Economics and Blockchain Architecture
The data availability wars of 2026 represent more than infrastructure competition—they're reshaping fundamental assumptions about how blockchains scale and how rollup economics work.
Celestia's Matcha upgrade and Fibre Blockspace roadmap make it clear that data availability is no longer the bottleneck for blockchain scalability. With 128MB blocks in production and 1 Tb/s demonstrated in testing, the constraint shifts elsewhere—to execution layer optimization, state growth management, and cross-rollup interoperability. This is a profound shift. For years, the assumption was that data availability would limit how many rollups could scale simultaneously. Celestia is systematically invalidating that assumption.
The modular architecture philosophy is winning. Every major rollup framework now supports pluggable data availability layers rather than forcing dependence on Ethereum mainnet. This architectural choice validates the core insight behind Celestia's founding: that monolithic blockchains forcing every node to do everything create unnecessary trade-offs, while modular separation allows each layer to optimize independently.
Different DA layers are crystallizing around distinct use cases rather than competing head-to-head. Celestia serves rollups prioritizing cost efficiency, maximum decentralization, and proven production scale. EigenDA appeals to Ethereum-native projects willing to accept DAC trust assumptions for higher throughput. Avail targets multichain infrastructure needing neutral coordination across ecosystems. Rather than a single winner, the market is segmenting by architectural priorities.
Data availability costs are trending toward zero, which changes rollup business models. As Celestia's block sizes grow and competition intensifies, the marginal cost of posting data approaches negligible levels. This removes one of the largest variable costs in rollup operations, shifting economics toward fixed infrastructure costs (sequencers, provers, state storage) rather than per-transaction DA fees. Rollups can increasingly focus on execution innovation rather than worrying about data bottlenecks.
The next chapter of blockchain scaling isn't about whether rollups can access affordable data availability—Celestia's Matcha upgrade and Fibre roadmap make that inevitable. The question is what applications become possible when data is no longer the constraint. High-frequency trading venues running entirely on-chain. Massive multiplayer gaming worlds with persistent state. AI model coordination across decentralized compute networks. These applications were economically infeasible when data availability limited throughput and spiked costs unpredictably. Now the infrastructure exists to support them at scale.
For blockchain developers in 2026, the data availability layer choice has become as critical as choosing which L1 to build on was in 2020. Celestia's market position, production-proven scalability roadmap, and ecosystem integrations make it the safe default. EigenDA offers higher throughput for Ethereum-aligned projects accepting DAC trust models. Avail provides multichain flexibility for teams coordinating across ecosystems. All three have viable paths forward—but Celestia's 50% market share, Matcha upgrade, and Fibre vision position it to define what "data availability at scale" means for the next generation of blockchain infrastructure.
Sources
- Introducing the Matcha upgrade: towards 128MB blocks - Celestia Blog
- Celestia Matcha Upgrade: Increased Block Size & Lower Inflation - IndexBox
- Choosing Your Data Availability Layer - Celestia, Avail, and EigenDA Compared - Eclipse Labs
- L2 Data Availability Layer: A Comparison of Celestia, EigenDA, and Avail - Technorely
- Introducing Fibre: 1Tb/s of blockspace - Celestia Blog
- Celestia Unveils Vision 2.0, Targets 1Tbps Blockspace for Global Markets - Blockchain News
- Avail DA launches on mainnet as native AVAIL token goes live - The Block
- Arbitrum, Optimism, Polygon, StarkWare and zkSync to integrate with Avail for data availability - The Block
- Celestia's Competitive Edge in Data Availability: A Deep Dive - BlockEden.xyz
- Layer 2 Adoption 2026 Predictions - Cryptopolitan