Celestia and the Modular Thesis: DA Layers, Blob Economics, and TIA in 2026

Celestia is the original sovereign data availability (DA) chain — a blockchain whose only job is to make raw transaction data publicly retrievable and provably available, while leaving execution and settlement to other layers. Since its mainnet launch in October 2023, Celestia has gone from research curiosity to a working layer underneath dozens of rollups, accelerating the modular thesis from whiteboard to production. This guide walks through what Celestia actually is, the data-availability-sampling math that makes it possible, blob economics, the four-way DA market (Celestia, EigenDA, Avail, PeerDAS), live production users, the modular-vs-monolithic narrative, TIA tokenomics, and the structural risks every DA consumer should understand.

What is data availability and why does it matter?

A rollup is, in essence, a chain that batches transactions, executes them off the L1, and posts a compressed summary back. For anyone to verify the rollup state — to challenge a fraud, to reconstruct an L2 wallet history, to exit if the operator vanishes — they need the underlying transaction data. If that data is published but unretrievable (an operator can claim it exists but not actually serve it), the rollup degrades to a permissioned database.

Historically, rollups posted call data to Ethereum L1, paying L1 gas prices for permanent storage. After EIP-4844 (March 2024) introduced blob space, rollups switched to blob posting — much cheaper, ~18 days of retention, with a separate fee market. But blob space on Ethereum is bounded (~6 blobs per block target, ~3 per block on average) and gets expensive when demand spikes.

DA layers like Celestia offer a different solution: a dedicated chain whose only job is to host blob-style data, with light clients that can verify availability without downloading everything. The trade-off is security — Celestia's security budget is smaller than Ethereum's — for cost — typically 10-100x cheaper than Ethereum 4844 blobs.

How Celestia works

Erasure coding and the 2D Reed-Solomon matrix

When Celestia receives a blob, it arranges the data in a 2D square matrix. Each row and each column is then erasure-coded with Reed-Solomon — extending the data so that any 25 percent of the resulting chunks can reconstruct the original. The matrix is committed to via row and column merkle roots, with a single overall data root anchored in the block header.

Why 2D? Because an attacker who tries to withhold data must hide enough chunks across both rows and columns simultaneously. With 2D coding, hiding 50 percent of the data requires hiding entire rows or columns — a structurally detectable pattern. If even one honest light client samples a missing chunk, the attack fails.

Data availability sampling

A light client doesn't download the full block. Instead, it picks N random chunks from across the matrix, requests them from peers, and verifies each against the block header. Each successful sample reduces the probability of undetected unavailability exponentially. With ~30 samples per block, a light client gets >99.99 percent confidence the data is fully retrievable — without trusting any individual full node.

This is the magic. As more light clients sample, security increases without anyone needing more bandwidth. DA security scales with the light-client population, not with the size of the data.

Namespaces

Celestia partitions every block into namespaces — logical lanes addressed by ID. A rollup writes its blobs tagged with its own namespace; downstream consumers (the rollup's full nodes, light clients, bridges) only fetch blobs from that specific namespace. This makes Celestia an efficient multi-tenant DA layer: a thousand rollups can share the same chain without each having to download one another's data.

The namespace ID is part of the blob's commitment, so a rollup can prove inclusion of its data in a Celestia block without needing the rest of the block's contents.

Consensus

Celestia uses a Tendermint-derived BFT consensus (CometBFT) with proof-of-stake validators. Block time is ~6 seconds (reduced from 12 seconds in the late-2024 Ginger upgrade), finality is single-slot (1 block confirmation). Validator set size is currently capped at 100 active validators, expanding over time. The economic security comes from staked TIA — currently ~$1-2B at active prices, a fraction of Ethereum's $100B+ but enough to make naive attacks expensive.

Blobstream

For Ethereum rollups using Celestia DA, the bridge is Blobstream — a contract on Ethereum that periodically receives Merkle-compressed proofs of Celestia data roots, signed by Celestia validators. An optimistic or ZK rollup settling on Ethereum can reference Blobstream proofs to assert 'my batched data lives at Celestia data root X.' Blobstream is the connective tissue between Celestia DA and Ethereum settlement.

Blob fee economics

Celestia charges a gas-per-byte fee for blob inclusion, with a base fee that adjusts via an EIP-1559-style mechanism: target block utilization (~70 percent), surge upward when demand exceeds, decay when below. Blob fees flow to validators as part of block rewards.

In early 2026, typical Celestia blob fees are well below $0.01 per kilobyte, often two orders of magnitude cheaper than Ethereum 4844 blob space at peak. A high-throughput rollup posting ~100 MB/day pays a few hundred dollars per day to Celestia versus tens of thousands on Ethereum.

The long-term question: as PeerDAS lands on Ethereum and 4844 blob throughput climbs 5-10x, the cost gap narrows. Celestia's bet is that demand grows faster than Ethereum DA capacity, leaving room for both.

The four-way DA market

Celestia

Sovereign DA chain. Own consensus, own validators, own token. Strengths: independent from Ethereum, cheap, mature DAS implementation, broad ecosystem support. Weaknesses: smaller security budget, requires Blobstream bridge for Ethereum-settled rollups.

EigenDA

A DA service built on EigenLayer restaking. Operators run dispersers and signers; security comes from restaked ETH (~$15-20B in restaked assets as of early 2026 backing EigenDA among other AVSs). Strengths: Ethereum-aligned security, deep institutional adoption (notably Eigen Labs ecosystem rollups), no separate token to acquire for posting (denominated in ETH-equivalent fees). Weaknesses: more centralized disperser architecture, no permissionless DAS in v1 (relies on attestations from a curated operator set), security budget legible but more abstract.

Avail

Polygon Labs spinout, separate token (AVAIL launched in 2024). Architecture parallels Celestia — KZG-based DAS, namespaces, validator set on substrate. Strengths: tight integration with Polygon CDK rollup stack, strong distribution into Polygon-native rollups. Weaknesses: smaller economic security, later launch than Celestia.

Ethereum PeerDAS

Ethereum's native roadmap. PeerDAS extends 4844 with peer-to-peer column sampling, letting nodes verify data availability without downloading the full blob set. PeerDAS activated on Ethereum mainnet via the Fusaka hard fork in December 2025, though its full blob-throughput increase phases in through 2026 via subsequent parameter (BPO) forks. When PeerDAS is live and stable, Ethereum 4844 blob throughput could grow 5-10x with the same node bandwidth, dramatically lowering blob fees and reducing the economic case for external DA.

The steady-state guess: cost-sensitive rollups use Celestia or EigenDA; security-maximizing rollups stay on Ethereum 4844/PeerDAS; the market segments by the rollup's risk tolerance.

Production users

By Q1 2026, more than 60 chains use Celestia DA in production. Notable categories:

Sovereign rollups — Manta Pacific (settled on Ethereum, DA on Celestia), Lyra Finance, Movement Labs Move-VM rollups.
Cosmos zones and chainlets — Stride zone, Dymension RollApps, Saga's chainlets, several Cosmos sovereign chains.
Rollup-as-a-service deployments — Caldera-deployed rollups, Conduit-deployed rollups, AltLayer optimistic rollups, all using Celestia as a default DA option.
Bitcoin DA-curious projects — multiple BitVM-style and sovereign-Bitcoin-rollup projects testing Celestia for transaction-data publication.

Daily blob throughput crossed 10-15 GB consistently in 2025 and hit 30+ GB on peak days in early 2026 as more rollups onboarded.

Modular vs monolithic

The broader narrative: monolithic chains (Solana, base-layer Ethereum, Aptos) bundle every layer into one stack and try to scale by making the one chain bigger. Modular chains decouple — execution, settlement, DA, and consensus can each be a different chain — and rely on composability to scale.

Monolithic strengths: low latency, unified state, simple developer mental model, no bridging risk between layers. Modular strengths: each layer can specialize, DA scales separately from execution, the same DA layer can serve thousands of rollups.

In 2026, Solana represents the strongest monolithic case: 65,000+ TPS theoretical, sub-second finality, atomic composability. The modular case is the entire Ethereum-aligned rollup ecosystem plus Cosmos sovereign chains. Both are growing; neither has knocked the other out. The likely future: a small number of monolithic giants (Solana, Aptos/Sui) plus a sprawling modular ecosystem (Ethereum L2s, Celestia, EigenDA, Bitcoin rollups, Cosmos zones).

TIA tokenomics

Initial supply: 1,000,000,000 TIA at mainnet (Oct 31, 2023).
Inflation: started at 8 percent year 1, decays 10 percent per year, asymptotes near 1.5 percent. Current ~6 percent in 2026.
Use cases: pay blob fees, stake to validate, delegate to validators, governance.
Airdrop: ~60M TIA distributed Nov 2023 to ~580K addresses across Ethereum rollup users, Cosmos stakers, IBC participants, and developers. One of the largest single-day wealth distributions in crypto.
Unlocks: investor and team tokens unlocked through 2025 (a continuing overhang); by mid-2026, most early supply is unlocked, simplifying the supply picture.
Burn vs issuance: blob fees flow to validators (no native burn); inflation continues regardless of demand. The structural concern: TIA is more inflationary than burning fees offset, making price appreciation rely on demand-side growth.

Risks

Security budget gap. Celestia's ~$1-2B economic security is dwarfed by Ethereum's ~$100B+. A motivated attacker with deep capital could in theory overwhelm Celestia's validator set; same attacker would not even attempt Ethereum.

Bridge risk. Blobstream is the trust assumption for Ethereum-settled rollups using Celestia. A bug in Blobstream could let bad data roots be attested or good ones be censored. Multiple audits have hardened the implementation, but it remains a critical surface.

DAS population assumption. Data availability sampling security degrades if too few light clients sample honestly. Celestia must continually grow its light-client population; failures here weaken the security claim.

Inflation overhang. TIA's structural inflation (~6 percent) without compensating fee burn creates persistent supply pressure. Demand growth must outpace issuance for token price to hold.

Competition from PeerDAS. If Ethereum native blob throughput grows 10x with PeerDAS, the cost gap that drives external DA adoption narrows materially. Celestia would need to either innovate further (e.g., much higher throughput, novel DA primitives) or accept a smaller share of Ethereum-aligned rollup demand.

Operational maturity. Celestia is younger than Ethereum (2 years vs 10). Production incidents have been minor so far, but the operational track record under stress is shorter than the alternatives.

Bottom line

Celestia turned the modular thesis from research paper to production reality. The DA-only chain is now an active layer underneath dozens of rollups, processing tens of GB of blob data per day at fees an order of magnitude below Ethereum's. Whether Celestia retains share long-term depends on three variables: PeerDAS execution speed on Ethereum, EigenDA's institutional pull, and Celestia's own ability to scale throughput while keeping security credible. As of 2026 it remains the cleanest sovereign DA play in the market and the de facto standard for non-Ethereum-aligned rollup ecosystems. TIA, meanwhile, is a pure bet on DA demand growth outpacing 6 percent inflation — the cleanest such bet in crypto.

Compare: Arbitrum vs Optimism
Compare: ARB vs OP
Compare: ATOM vs DOT

Sources and further reading

Celestia whitepaper — https://celestia.org/papers/Celestia-Whitepaper.pdf
Mustafa Al-Bassam, Alberto Sonnino, Vitalik Buterin (2018), Fraud and Data Availability Proofs
Celestia documentation — https://docs.celestia.org
L2BEAT DA risk framework — https://l2beat.com/data-availability
EigenDA technical brief — https://docs.eigenda.xyz
Ethereum PeerDAS spec — https://github.com/ethereum/consensus-specs

About the author

DeFi Intel Research is the in-house research team at DeFi Intel, focused on on-chain capital markets, MEV, ZK infrastructure, and verifiable AI.

Last updated: 2026-04-30

Entities mentioned

Celestia EigenDA Avail Ethereum

Frequently asked questions

What is Celestia in plain English?

Celestia is a data availability (DA) layer — a blockchain whose only job is to publish raw transaction data and prove that data is actually available to anyone who wants to download it. Rollups and other chains pay Celestia to store their batched data instead of posting it to Ethereum or running their own consensus. The pitch: cheaper data, faster finality for L2s, and a clean modular split between execution (where the rollup runs), settlement (where state proofs are verified), and DA (where data lives).

How does Celestia actually work — what is data availability sampling?

When a rollup posts data to Celestia, the network breaks the data into a 2D matrix and applies erasure coding so any 25 percent of the original chunks can reconstruct the whole. Light nodes then sample tiny random fragments from each block; if enough samples succeed, the network statistically guarantees the full data is available without any single node downloading it all. This is data availability sampling (DAS), and it's the unlock — light clients gain near-full-node security on data availability, which lets DA scale with light-node count rather than requiring everyone to be a beefy full node.

What are namespaces and how does blob pricing work?

Celestia organizes posted data into namespaces — distinct logical lanes, one per rollup or application. A rollup writes its blob (a chunk of transaction data) tagged with its namespace; light clients on that rollup only download blobs from the relevant namespace, not the whole Celestia block. Pricing is gas-per-byte: posters pay TIA proportional to blob size, and a base fee adjusts dynamically based on block-space demand (similar to EIP-1559). As of early 2026, Celestia's blob fee is a tiny fraction (often 100x cheaper) of equivalent Ethereum 4844 blob space for the same volume.

How does Celestia compare to EigenDA, Avail, and Ethereum's PeerDAS?

Four flavors of the same problem. (1) Celestia — sovereign DA chain with its own consensus and DAS, namespaces, TIA-paid blobs. Live since Oct 2023. (2) EigenDA — DA service built on EigenLayer restaking; Ethereum-aligned, lower-cost, but security comes from restaked ETH and a centralized disperser. Live since 2024. (3) Avail — Polygon Labs spinout with similar DAS architecture to Celestia, tightly integrated with the Polygon CDK rollup stack. (4) PeerDAS — Ethereum's own roadmap to data-availability sampling, which landed via the Fusaka hard fork in December 2025, dramatically expanding 4844 blob throughput. PeerDAS is the long-term competitor; until it ships at scale, external DA layers (Celestia, EigenDA) own the cost-sensitive segment.

Who actually uses Celestia for DA?

Production users span ecosystems. (1) Cosmos rollups — Stride zone, Dymension's RollApps, Saga's chainlets. (2) Ethereum rollup stacks — Manta Network's Pacific (Celestia for DA, Ethereum for settlement), Caldera's rollup-as-a-service deployments, Movement Labs' Move-VM rollups, several OP-Stack and Arbitrum Orbit chains. (3) Bitcoin rollups — multiple BitVM-style rollups using Celestia as DA. By Q1 2026, more than 60 chains have integrated Celestia DA, and Celestia routinely settles 10-30 GB of blob data per day, an order of magnitude more than it did at launch.

What is the modular vs monolithic debate?

Monolithic chains (Solana, classic Ethereum, BSC, Aptos) bundle execution, settlement, consensus, and data availability into one stack — fast and unified, but scaling means making one chain bigger. Modular chains separate the layers and let each scale independently — Celestia for DA, Ethereum for settlement, the rollup itself for execution, optionally a third layer for proof aggregation. The modular pitch: composable specialization. The monolithic counter-pitch: simplicity, lower latency, and unified state. Reality in 2026: the line is blurring. Solana stays committedly monolithic. Ethereum is structurally modular (rollups + 4844). Celestia is the modular pure-play DA. The race is whether monolithic chains can scale execution faster than modular stacks can compose.

What does TIA do and how is the token positioned?

TIA pays for blob space (every namespace post burns or pays TIA), secures the network via proof-of-stake (validators stake TIA, delegators delegate), and governs protocol parameters. Initial supply was 1 billion at the Oct 2023 mainnet launch. Inflation started at 8 percent annually and decreases by 10 percent per year (current rate is roughly 6 percent in 2026), targeting a long-run rate of around 1.5 percent. The TIA airdrop in Nov 2023 was one of the largest in crypto history, distributing ~60M tokens to ~580K addresses including Ethereum rollup users, Cosmos stakers, and IBC participants — generating multi-thousand-dollar payouts for engaged users.

What are the risks of using Celestia DA?

Six live risks. (1) Security via a smaller validator set than Ethereum — Celestia's economic security (~$1-2B staked) is far below Ethereum's ($100B+), so DA bridge cost-of-attack is lower. (2) Bridge risk — using Celestia DA from an Ethereum-settled rollup requires a DA attestation bridge (Blobstream); a buggy bridge could make Celestia data unverifiable on Ethereum. (3) DAS assumption risk — DAS security depends on enough light clients sampling honestly; a thin light-client population reduces guarantees. (4) Token-fee dependency — if TIA price drops, validator rewards in USD terms shrink, potentially destabilizing security. (5) Competition — EigenDA and PeerDAS both target Celestia's market; PeerDAS in particular could collapse blob fees on Ethereum and undercut external DA. (6) Inflation overhang — 6+ percent issuance with limited fee burn means TIA is structurally inflationary unless blob demand catches up.

Stay current on this topic

Get the weekly DeFi Intel brief — entity-graph intelligence delivered weekly, free to your inbox.

Subscribe free See Pro →