How did data availability — the cheapest, most boring layer of the rollup stack — turn into the most aggressively priced fight in crypto infrastructure?
October 31, 2023: Celestia Goes Live and the Modular Thesis Stops Being a Whitepaper
Let me concede something upfront. Celestia did the thing nobody else in the modular DA conversation had actually done at scale — they shipped a sovereign data-availability layer with its own consensus, its own token economics, and a working namespaced commitment scheme. Mainnet beta went live on October 31, 2023. The TIA airdrop landed the same day. Whether you liked the design choices or not, the launch was real, the chain was producing blocks, and rollups could start posting blobs to it within hours.
This matters for the pricing story because Celestia's launch reset the conversation. Before October 2023, "alt-DA" meant a centralized data committee or a roll-your-own validator set running off a Tendermint fork. After October 2023, alt-DA meant a chain you could verify, a token you could price, and a fee market you could compare directly to Ethereum calldata.
The Crypto Twitter framing at the time was that Celestia would "eat Ethereum's data fees." That framing was wrong in a specific way — Ethereum had not yet shipped blobs, and the comparison was Celestia versus Ethereum calldata, which is roughly the same comparison as pricing electricity against candles. Useful to nobody once the lights came on.
But the analytical fact stands. Celestia was the first independently-secured DA layer where you could pull the fee data from a public explorer and actually price it. Everything that follows in this timeline only makes sense if you start the clock here.
March 13, 2024: Ethereum Ships Blobs and Resets the DA Floor
The Dencun upgrade activated on March 13, 2024. The headline change, EIP-4844, introduced blob-carrying transactions — a separate fee market specifically designed to make rollup data posting cheap.
This is the moment the alt-DA pricing thesis got harder to defend, and the moment it got more interesting. Pre-Dencun, the case for Celestia was "Ethereum calldata is expensive and Celestia is not." Post-Dencun, the case for any alt-DA layer became "Ethereum blobs are cheap, and we need to be cheaper still while offering something Ethereum cannot."
I want to be precise about what I can and cannot verify. The exact per-blob fee history sits on the public chain — anyone can pull it from a blob explorer or the standard Etherscan blob view. The base fee mechanic for blobs is its own EIP-1559-style market, which means rollup data costs swing wildly with congestion. I have looked at the public dashboards. I am not going to quote a specific cents-per-megabyte figure in this article because the figure I pulled this morning will be wrong by the time you read this. That is the actual story.
What changed permanently in March 2024 is the floor. Any DA layer that wants rollup business now has to argue against a blob market that is, on average, very cheap and getting cheaper as more rollups share blob space. The price war did not start with Celestia and EigenDA fighting each other. It started with both of them having to undercut a base layer that suddenly stopped being expensive.
April 9, 2024: EigenDA Mainnet and the Restaking Bet on DA
EigenDA went live on Ethereum mainnet on April 9, 2024. The architectural pitch was different from Celestia's. EigenDA does not run an independent consensus chain — it runs as an AVS (actively validated service) on top of EigenLayer's restaked ETH security model. Operators run DA nodes, erasure-code the data, sign attestations, and a smart contract on Ethereum verifies that the threshold of signatures has been collected.
The pricing implication is that EigenDA's cost structure looks different from a chain. There is no native gas token you pay in. There is no block-by-block fee auction. Instead there is a service fee paid to operators, denominated in ETH, with an economic security model that scales with restaked ETH rather than with the market cap of a native token.
I think this is the part of the EigenDA design that the marketing layer of crypto consistently undersells. The competitive question between EigenDA and Celestia is not "which has lower per-byte fees this week." The competitive question is which security model rollup operators trust enough to host their canonical data, and which fee model their treasury can budget against without surprise. A chain-with-token model and a restaked-AVS model behave very differently in a bear market, and the analytical work nobody is doing publicly is stress-testing both under a 70% drawdown of their underlying collateral.
The fact that EigenDA launched five months after Celestia is itself a pricing signal. The market had a Schelling point. Whatever EigenDA was going to charge had to be visibly cheaper than Celestia's effective rate at launch, or the modular DA market would coalesce around the incumbent.
Through 2024: Rollup Migrations Begin and the Fee Math Becomes Public
Through the second half of 2024, the rollup ecosystem started doing something that, frankly, the analytical infrastructure of crypto was not ready for. Rollups began publicly migrating between DA layers. Some moved from Ethereum calldata to blobs. Some moved from blobs to Celestia. A smaller number announced EigenDA integrations. Each migration generated a Twitter thread, a blog post, a Dune dashboard, and a set of fee comparisons that were almost never apples-to-apples.
The reason the comparisons were not apples-to-apples is that DA pricing has at least four moving parts. There is the raw per-byte fee. There is the security premium — you are not just paying for bytes, you are paying for the assurance that those bytes are retrievable and attestable. There is the bridge cost — if your DA layer is not the same chain as your settlement layer, you are paying for a light client or a bridge committee. And there is the latency cost, which shows up as either delayed finality or as a separate fast-confirmation service.
When I priced what each layer "actually charges a rollup" the calculation only made sense once I had all four numbers and a usage profile to weight them by. A high-throughput perp DEX is paying very differently from a low-throughput governance rollup, even when the headline per-megabyte rate is the same. The migrations that happened in late 2024 made this brutally obvious, because some rollups switched DA providers and then switched back within a quarter once the real bill came in.
This is where the "price war" framing starts being honest. By late 2024, both Celestia and EigenDA were pitching rollups individually, and the published rate cards were not the rates anyone was paying. The published rates were the anchor. The actual rates were negotiated, often for committed throughput.
2025: The Economic Squeeze and the Question Nobody on Crypto Twitter Asks
Through 2025 the pricing pressure on alt-DA has only intensified, because the blob market on Ethereum has done exactly what its designers hoped. More rollups, more blob consumers, more efficient packing, lower average cost. Every time Ethereum's blob fees compress, both Celestia and EigenDA have to compress further to maintain their relative advantage.
The question the price-war framing obscures is whether either layer can be sustainably profitable at the rates required to win rollup business in a fully blob-saturated world. I do not think the answer is obvious. Celestia's economic model depends on TIA having a market cap that supports the security budget of the chain, and TIA's market cap is correlated with rollup adoption of Celestia, which is in turn correlated with how cheap Celestia is, which is in turn what compresses TIA's fee revenue. The loop is fragile. EigenDA's model depends on restaked ETH being abundant and willing to underwrite DA-specific slashing risk for a yield denominated in service fees that are themselves being competed down. Also fragile.
For perspective on the scale we are talking about, Binance does roughly $18.5 billion in daily exchange volume according to current disclosures. The entire combined daily revenue of Celestia and EigenDA, based on the public dashboards I have looked at, is a vanishingly small fraction of that. This is a price war in a market that, right now, prices in the low single-digit millions per day at most. The capital looking at this fight is much larger than the fees being fought over.
What It All Means
I think the pricing war between EigenDA and Celestia is real, but I do not think it is the war either team's marketing wants you to think it is. Both teams talk about per-byte fees because that is the simplest comparison to put on a slide. The actual competition is for which architecture the next generation of rollups commits to for the security horizon they are planning against — three to five years out, not three to five months.
The honest analytical move is to treat the published rate cards as the floor of the conversation, not the substance of it. Anyone integrating either DA layer at meaningful throughput is negotiating, and the published numbers do not survive contact with a serious procurement conversation. This is not unique to DA — it is how every infrastructure market in crypto eventually behaves once the buyers get sophisticated enough to ask for SLAs.
What you should actually do, if you are pricing DA for a rollup you are building or evaluating, is construct the four-component cost model I sketched above — raw per-byte, security premium, bridge cost, latency cost — and run it against your expected usage profile for at least two epochs of blob market congestion. Not the average. The 95th percentile. The DA cost that breaks your rollup's unit economics is not the mean cost. It is the cost on the day the blob market spikes and your team has not budgeted for the spike.
Watch three things over the next two quarters. First, the spread between published rate cards and the rates that new rollup integration announcements actually disclose — that spread is the negotiation premium and it tells you who has pricing power. Second, the ratio of EigenDA service fees to the underlying restaked ETH security budget — if that ratio collapses, the AVS model is being subsidized, and subsidies in crypto have a known half-life. Third, the rate at which any rollup that migrated to either Celestia or EigenDA in 2024 is publicly defending the decision in 2026 — silence is signal. The rollups that quietly leave are the ones that already priced the thing properly.
FAQ
How much does it actually cost to post data on EigenDA or Celestia today?
I am deliberately not quoting a single per-megabyte figure here, because the rates move daily and the published rate cards are not the rates large integrators pay. Pull the numbers yourself from each protocol's public dashboard and the relevant block explorers before you cite anything — the figures rotate too quickly for any static article to be honest about them. The four-component cost model in the body matters more than the spot rate.
Is Celestia or EigenDA cheaper for a high-throughput rollup?
It depends on whether you are pricing the raw per-byte fee or the loaded cost. Once you add bridge costs, latency assumptions, and the security premium each layer charges, the ranking flips depending on the rollup's usage profile. A perpetuals DEX with bursty traffic and tight latency requirements prices differently from a low-throughput application chain. Anyone telling you one is unambiguously cheaper has not actually built the model against a realistic usage curve.
Did Ethereum blobs make alt-DA layers obsolete?
No, but they rebased the conversation. Pre-Dencun, alt-DA was competing against expensive calldata. Post-Dencun, it is competing against a deliberately cheap blob market that gets cheaper as more rollups share blob space. The case for an alt-DA layer is no longer "blobs are expensive" — it has to be materially cheaper at scale, materially faster, or offering a sovereignty property Ethereum blobs do not.
Why does EigenDA not have a native token like TIA?
Architecturally EigenDA is an AVS on EigenLayer, not an independent chain. Its security is sourced from restaked ETH and its fees are denominated in ETH paid to operators. There is no native gas token because there is no native chain — the attestations settle on Ethereum and the security model inherits from restaking. That is the design choice, not an oversight, and the pricing implications run all the way through the economics.
What is the actual risk if a rollup picks the wrong DA layer?
The risk is not that you cannot get your data back tomorrow — both layers have working retrieval. The risk is that the DA layer you picked compresses its fees so aggressively that it cannot fund its own security budget, at which point you have a slow-motion solvency question on the layer holding your canonical state. Migrating off a DA layer is operationally expensive. The cost of being wrong is the migration cost, not the daily fee delta.
How does this DA market compare in scale to centralized exchanges?
By several orders of magnitude smaller. Binance reports roughly $18.5 billion in daily exchange volume on current disclosures. The combined daily DA fee revenue across Celestia and EigenDA, based on public dashboards, is a tiny fraction of that. The narrative attention on the DA price war substantially exceeds the dollar value being competed over today — which is precisely why the long-run structural bet matters more than the spot rate.
Should I trust the published rate cards from either protocol?
Treat them as the public anchor, not the negotiated price. Any rollup integrating at meaningful throughput is having a procurement conversation that does not reach Twitter, and the rate that ends up in the contract is rarely the rate on the public page. The spread between the two is itself the most interesting datapoint, because it tells you who has pricing power in any given quarter.