PeerDAS increases blob capacity via sampling

PeerDAS uses 1D Reed-Solomon erasure coding to expand blob data. It treats data as polynomial coefficients and evaluates the polynomial at additional points to create 2N pieces of data. If a malicious proposer hides one byte, they must hide more than half of the extended data to prevent reconstruction. Nodes use KZG cell proofs to verify a sampled piece against the blob commitment. The PeerDAS upgrade changes how nodes handle blob data. Instead of downloading everything, nodes subscribe to specific subnets consisting of 128 columns. A typical validator handles 8 of these subnets. This prevents storage requirements from growing linearly with the number of blobs. High-balance validators with 4,096 ETH or more act as supernodes to ensure the network can reconstruct all data.

The network reorganizes around 128 column subnets. A node handles only a small sample of the data, which reduces bandwidth requirements. For 8 subnets, a node handles 16 KiB of data per slot of blob payload at current targets. You know the fundamentals of L2 scaling, so the shift from full download to sampling explains the reduction in costs.

L2 fees dropped after BPO deployments

L2 transaction fees fell to the $0.005 to $0.02 range by Q1 2026. This drop follows the Fusaka upgrade on 3 December 2025 and subsequent BPO deployments. The implementation of BPO2 on 7 January 2026 raised the target and maximum blob counts to 14 and 21 respectively, which helped the Ethereum ecosystem manage the increased data without overwhelming the home stakers. On 14 February 2026, an Arbitrum swap cost $0.008. This was a significant reduction from the $0.34 cost in early 2025.

Optimism implemented the Delta upgrade to reduce fixed overhead. Zora-sepolia saw a 90% decrease in those costs. The L2 gas fee on Optimism follows a formula: Layer 1 Gas Fee = Fee Scalar L1 Gas Price (Calldata + Fixed Overhead). The update to EVM Equivalence on 11 November reduced the fixed overhead from 2750 to 2100 gas per transaction. This change, with a Fee Scalar reduction from 1.5x to 1.24x, lowered fees by 30%. Rollup costs make up about 99.6% of a transaction fee on OP Mainnet.

Ethereum competes with Celestia on data availability

Celestia provides a modular alternative by separating consensus and data availability. It uses Data Availability Sampling to allow light nodes to verify data with minimal resources. This architecture allows multiple execution layers to operate concurrently, which makes it easier to handle high transaction volumes. For developers prioritizing cost, Celestia is a strategic bet. For those prioritizing security, Ethereum is the conservative choice.

Ethereum leverages the economic security of its $100B staked value. Celestia allows connected chains to maintain governance while relying on a shared base layer for consensus. When Ethereum blobs cost $0.50, the trade-off for using Celestia was easy to defend. When they cost a fraction of a cent, users must argue harder for why they should accept weaker security.

Validator latency remains a problem for some

The increase in blob counts causes a linear increase in latency. For 95% of validators, PeerDAS improved validation latency compared to Pectra. However, the unluckiest 1% of validators experience worse block latency. When the count hits 9 blobs, the p99 latency exceeds 5 seconds. This value reaches 12.4 seconds when the count hits 15. Failed attestation rates also rise as blobs increase. For 15 blobs, the p95 failed attestation rate reaches 7%. The unluckiest 1% of validators see p99 values that exceed the 4 second deadline. This indicates the network faces physical limits. Does the increasing number of blobs eventually make PeerDAS unreliable for all validators?