The tightening squeeze on Celestia’s modular dominance

Celestia faces a difficult balancing act as it manages its transition toward a sustainable blob economy. After the adoption of upgrade v6, CIP-41, TIA inflation dropped to 2.5% in November 2025. This inflation decreases by 6.7% every year until it reaches the 1.5% floor. While the network supports 55 networks and 4,100 GB of data, the proposal to replace TIA issuance with protocol revenue remains in a feasibility phase. This plan attempts to introduce paid capacity commitments and bonded services to generate revenue beyond basic pricing. I find the reliance on these unapproved fee systems a risky gamble for stakers expecting stable rewards. The proposal moves from a $225,000 feasibility phase to an $825,000 implementation phase, eventually leading to a $450,000 independent review. Because the Ethereum Pectra upgrade in May 2025 doubled blob throughput, blob object rollback fees dropped by almost 100% from $16,250 per day. This massive drop proves that increasing throughput often results in lower profit per transaction, which impacts Celestia’s pricing strategy. Celestia reported a 30-day average DA cost of $0.0188 per MB on September 22, compared to $0.0325 for Ethereum blobs.

EigenDA presents a direct challenge to Celestia’s cost leadership through its high throughput and Ethereum-backed security. The Zeeve partnership with EigenDA provides rollups with decentralized data availability solutions where the combination allows developers to launch rollups with just a few clicks while they also benefit from high-uptime sequencing and provisioning without any manual configuration. EigenDA reaches 15 MB/s of bandwidth, which exceeds the 0.2 MB/s capacity of Ethereum’s current blob structure.

I find the EigenDA cost model aggressive because it leverages restaked ETH to provide security without the constraints of the Ethereum mainnet. EigenDA uses erasure coding to split data into small chunks, meaning nodes only download tiny parts of the data. This mechanism ensures that even if half the nodes leave, the system remains unaffected because the full data state can still be reconstructed. EigenDA currently has 115 operators and over 42,000 stakers delegating over 400,000 ETH to secure the network. Because the design allows for horizontal scalability, throughput increases as more operators join. If you are building a high-throughput rollup, do you prefer the higher cost of EigenDA or the lower pricing of Celestia?

The competition between these protocols reveals a tension in how modular networks scale. Celestia relies on light nodes to perform data availability sampling with 99% confidence using only 15 samples. This method uses 2D Reed-Solomon erasure encoding to expand block size, which allows the network to scale as more clients join. In a block with 4,096 shares and a code rate of 1/4, only 0.4% of the block, or 15 shares, must be downloaded to reach 99% confidence. A single row or a single column plus 2N Merkle roots must be downloaded for a sub-maximal guarantee.

EigenDA operates differently by breaking calldata into small chunks and using KCG polynomial commitments to facilitate a system where each node only downloads a small chunk. This approach allows EigenDA to process data at a multiple of 200x times that of Ethereum mainnet. I find the lack of a default safe bridge to Ethereum from EigenLayer a significant problem for developers seeking easy integration. Celestia light nodes do not verify transaction data because they only verify consensus and DA. In Celestia, the size of each block can increase as more light clients perform sampling requests. This increase in block size aids scalability, as the security of the network grows with the number of clients. Celestia uses a namespaced and ordered Merkle Tree to allow applications to query transactions associated only with their namespace ID. The hash of each non-leaf node includes the leftMinNs, leftMaxNs, and the hash of x, which allows nodes to determine the range of namespaces the subtree holds messages for.