PeerDAS and the Future of Ethereum: Transforming Data Availability and Layer 2 Economics
Ethereum validators used to download every byte of blob data posted to the network — roughly 750 MB per day and climbing. Since December 3, 2025, they don't have to. The Fusaka hard fork introduced PeerDAS (Peer Data Availability Sampling), a cryptographic technique that lets nodes verify data availability by checking only a small random slice instead of the entire payload. Three months in, the results are reshaping the economics of every major Layer 2.
What PeerDAS Actually Changes Under the Hood
PeerDAS, formalized as EIP-7594, replaces Ethereum's "everyone downloads everything" model with probabilistic verification through Reed-Solomon erasure coding. Here's how it works in practice:
- Blob extension: Each blob's data is extended using 1D erasure coding, doubling its size but adding redundancy that makes reconstruction possible from any 50% of the pieces.
- Column distribution: The extended data is split into 128 columns. Each regular node subscribes to at least 8 randomly chosen column subnets — meaning it downloads roughly 1/16 of the extended data, or 1/8 of the original.
- Supernode enforcement: Nodes backing validators with a combined stake of 4,096 ETH or more must subscribe to all 128 column subnets and act as data healers, filling gaps for the rest of the network.
- Sampling verification: Any node can verify data availability by requesting a few random columns and checking them against KZG polynomial commitments, achieving a strong probabilistic guarantee without downloading the full dataset.
The net effect: validator bandwidth requirements drop by approximately 85%. A node that previously pulled 750 MB of blob data per day now handles around 112 MB while maintaining the same security guarantees.
The Blob Throughput Ramp-Up
Fusaka didn't just flip a switch. Ethereum's core developers introduced a novel mechanism called Blob Parameter Only (BPO) forks via EIP-7892, allowing blob limits to increase through lightweight hard forks that don't require bundling with major named upgrades.
The scaling has been deliberately phased:
| Fork | Date | Blob Target | Blob Maximum | Data per Block |
|---|---|---|---|---|
| Pre-Fusaka | Before Dec 2025 | 6 | 9 | ~768 KB |
| Fusaka + PeerDAS | Dec 3, 2025 | 6 (with DAS) | 9 | ~768 KB |
| BPO1 | Dec 9, 2025 | 10 | 15 | ~1,920 KB |
| BPO2 | Jan 7, 2026 | 14 | 21 | ~2,688 KB |
| BPO3/BPO4 (planned) | 2026 | Up to 128 | TBD | ~16 MB |
Each blob holds 128 KB of data. At the current BPO2 parameters, Ethereum processes up to 2.7 MB of blob data per block. The long-term target of 128 blobs per block would push this to 16 MB — roughly twice what Celestia's mainnet currently offers.
L2 Fee Collapse: The Numbers That Matter
Data availability accounts for approximately 90% of Layer 2 operating costs. When you reduce DA costs by an order of magnitude, the impact on end-user fees is dramatic.
Before Fusaka, L2 transactions on Arbitrum, Optimism, and Base typically cost $0.05–$0.15 for a simple swap. Post-BPO2, those same transactions have stabilized around $0.004–$0.02, representing a 70–95% reduction stacked on top of the savings already delivered by the Dencun upgrade in March 2024.
Specific network impacts observed in early 2026:
- Arbitrum: 40–60% fee reduction within the first month post-Fusaka, with average swap costs dropping below $0.01
- Optimism and Superchain: The Superchain ecosystem immediately benefited from expanded blob capacity, with OP Mainnet transactions settling for $0.005–$0.01
- Base: As Coinbase's L2 with the highest transaction volume, Base saw the most dramatic throughput improvement, processing more transactions at lower per-unit costs
The trajectory toward 128 blobs per block suggests that L2 fees could approach the sub-$0.001 range by late 2026 — territory that makes micropayments and high-frequency DeFi strategies economically viable for the first time.
The DA Layer War: Does PeerDAS Make Celestia Obsolete?
When Fusaka shipped, the immediate question was whether Ethereum's native DA improvements would kill third-party data availability layers like Celestia, EigenDA, and Avail. Three months later, the answer is nuanced.
The case for Ethereum DA dominance: At 128 blobs per block, Ethereum would offer roughly 16 MB per block of DA capacity — double Celestia's current throughput. For rollups that already settle on Ethereum, using native blobs eliminates bridge risk and simplifies architecture. The security model inherits Ethereum's full validator set rather than relying on a separate consensus mechanism.
The case for continued DA specialization: Even at BPO2 parameters, Ethereum L2s pay significantly more per megabyte than Celestia users. Eclipse reported paying Celestia $0.07 per megabyte compared to $3.83 per megabyte for Ethereum blobs — a 55x cost advantage. This gap will narrow as blob capacity increases, but cost-sensitive rollups and high-throughput chains (gaming, social) may continue to prefer dedicated DA layers.
EigenDA's middle path: EigenDA V2 achieved 100 MB/s throughput by leveraging EigenLayer's restaking infrastructure, offering Ethereum-aligned security without competing for blob space. Its Data Availability Committee model trades some decentralization for raw performance, positioning it as the enterprise choice for Ethereum-native projects that need more throughput than blobs can provide.
The likely outcome is market segmentation: Ethereum blobs for security-maximizing rollups, Celestia for cost-sensitive and sovereign chains, and EigenDA for high-throughput Ethereum-aligned applications.
From 1D to 2D: The Road to Full Danksharding
PeerDAS is a milestone, not the destination. The full Danksharding vision — named after researcher Dankrad Feist — requires upgrading from PeerDAS's 1D erasure coding scheme to a 2D coding approach that operates across the entire blob data matrix rather than individual blobs.
The difference matters for the scaling ceiling. With 1D coding, each blob is independently encoded and verified. With 2D coding, the entire block's blob data forms a matrix where both rows and columns are erasure-coded. This enables:
- Higher sampling efficiency: Nodes can verify the entire data matrix with fewer samples
- Greater fault tolerance: Data reconstruction becomes possible even with more aggressive column losses
- Theoretical throughput: Target of 1 GB/s data availability, representing a 30,000x increase over pre-Fusaka capacity
The full transition requires further research on proof generation costs, network propagation times, and the interaction between 2D DAS and proposer-builder separation. It remains on the Ethereum roadmap under "The Surge" but without a firm deployment date.
What Comes Next: Glamsterdam and Hegota
Ethereum's 2026 roadmap moves rapidly beyond Fusaka with two more named upgrades:
Glamsterdam (expected May–June 2026) shifts focus from rollup scaling to Layer 1 efficiency. Its two headline EIPs are:
- EIP-7732 (Enshrined Proposer-Builder Separation): Formalizes how block builders bid for blocks and how validators participate, reducing MEV-related centralization pressures
- EIP-7928 (Block-Level Access Lists): Requires builders to declare which state they'll access before execution, enabling parallel processing and better scheduling
Hegota (expected H2 2026) targets node efficiency through:
- Verkle Trees: A new data structure that dramatically reduces the proof sizes needed for state verification, potentially shrinking node storage requirements and enabling truly stateless clients
- Historical data management: Improvements to how long-term chain data is stored and pruned
Together, these upgrades address the complementary challenge to data availability: making Ethereum's base layer fast and efficient enough to serve as settlement infrastructure for thousands of rollups simultaneously.
What This Means for Builders
The practical implications for developers building on Ethereum and its L2 ecosystem are significant:
- Cost models have changed permanently: Applications that were economically impossible at $0.10 per transaction are viable at $0.005. Recalculate your unit economics.
- DA layer choice is a strategic decision: Evaluate whether Ethereum-native blobs, Celestia, or EigenDA best fits your security requirements and cost sensitivity. The answer differs by use case.
- Prepare for blob abundance: As BPO3 and BPO4 roll out through 2026, blob capacity will continue expanding. Design your data posting strategies to take advantage of falling costs rather than optimizing for scarcity.
- Watch Glamsterdam for L1 changes: ePBS will change block building dynamics, and block-level access lists may affect how transactions are ordered and executed.
The Fusaka upgrade and its PeerDAS implementation represent Ethereum's clearest statement yet that the rollup-centric roadmap is working. Three months of production data confirm that probabilistic data verification is not just theoretically sound but practically transformative — reducing costs, expanding capacity, and setting the stage for the full Danksharding endgame that could make Ethereum's data layer virtually unlimited.
BlockEden.xyz provides high-performance RPC endpoints and data infrastructure for Ethereum and leading Layer 2 networks including Arbitrum, Optimism, and Base. As the DA landscape evolves post-Fusaka, our infrastructure scales with it. Explore our API marketplace to build on the new economics of abundant data availability.