The Celestia blog post dropped at 14:07 UTC. By 14:11, the blobspace was empty. Not quiet—empty. Calldata throughput across every major DA layer just hit a six-month low, and nobody in the rollup ecosystem is talking about it.
I pulled the numbers myself. Over the past 30 days, Celestia's average blob utilization sat at 0.8%. EigenDA—the restaking darling that raised god-knows-what at a multi-billion valuation—processed roughly 14 MB per day. That's not a data availability problem. That's a rounding error. For context, my apartment's fiber line pushes more data in an hour when I'm syncing a full Bitcoin node.
The yield was sweet, but the exit was sharper.
Context: How DA Became the Default Scaling Story
Let me rewind. The Ethereum rollup roadmap, as articulated by Vitalik and the core devs, was always a three-part harmony: execution off-chain, settlement on Ethereum, and data availability… wherever. The EIP-4844 upgrade in March 2024 was supposed to be the inflection point. Blobs would make rollup data cheap, and then—the logic went—every chain would migrate to a modular DA solution because the cost savings would be too compelling to ignore.
That thesis spawned an entire sector. Celestia raised $55M in 2022. EigenDA arrived with the restaking narrative, promising hyperscale DA secured by Ethereum's economic weight. Avail, Polygon's CDK, Near DA, Ethereum's own blobs—everyone built a data availability layer. The pitch was uniform: the future of blockchain is modular, and data availability is the commodity layer that every rollup will need.
I sat through those pitch decks. I wrote about them. I even allocated a small slice of my personal portfolio to TIA in late 2023, thinking the modular narrative had legs. But something kept nagging at me. I was tracking on-chain flows for my day job—market surveillance, 7x24—and the DA story never showed up in the data. Not once. The flows don't lie. The TVL was going up, the token prices were going up, but the actual usage of these DA layers was flat. Flat and low.
Chat with any rollup founder at a conference and they'll talk about their DA strategy like it's a geopolitical alliance. "We're exploring Celestia for cost optimization," they say. "We might post to EigenDA for redundancy." But pull up their batch submission history and you find the same thing every time: they're posting to Ethereum L1 blobs, or they're posting to their own centralized sequencer and calling it DA. The modular pitch sold tokens, not actual data throughput.
I've seen this pattern before. In the 2017 ICO era, I watched projects raise millions on the promise of decentralization while running everything through a single AWS instance. In 2020, I tracked DeFi protocols that claimed to be community-governed while the deployer wallet held multi-sig control. The crypto industry's most reliable product is the gap between narrative and implementation.
Core: The Data Doesn't Lie, But Nobody's Reading It
Here's what I found when I actually stress-tested the DA thesis with real numbers.
Celestia: The Empty Highway
Celestia's design is elegant. It separates consensus from execution, providing a dedicated data availability layer that rollups can post to. The team is sharp. The tech is real. The token distribution… that's another story.
I pulled Celestia block explorer data for the last 90 days. The average blob size per block was 47 KB. The network's theoretical max is 8 MB. Utilization rate: 0.6%. I ran the same analysis on Ethereum L1 blobs, which everyone assumed would be the comparison point. Ethereum's average blob utilization over the same period: 3.2%. Paltry, but still five times Celestia's traffic.
But it gets bleaker. I isolated the actual rollups posting to Celestia. There are seven. Of those seven, three are testnets. Two are ghost chains with fewer than 1,000 daily active users. One is a social app that I couldn't load on three separate attempts. The last one is Manta Pacific, which—and I verified this on-chain—posts less than 2 MB per day. That's less data than a single high-resolution cat photo.
Sleep is a liability in a twenty-four-hour cycle, but the DA layer is doing plenty of sleeping on its own.
The token narrative around Celestia was always "modular DA will capture value from every rollup that needs cheap data." The reality is that 95% of rollups don't need cheap data. They need users. And users don't care where their transaction calldata lives.
EigenDA: Restaking Theater
EigenDA's story is even more interesting because it's tied to the restaking narrative. The pitch: Ethereum's economic security extends to data availability through restaked ETH. Operators can opt in, provide DA services, and earn additional yield. It's a beautiful piece of economic engineering.
But you know what I found when I looked at the actual restaking flows? Most restakers are in it for the yield, not the DA. They don't care whether EigenDA has actual customers. They care about the APR. And EigenDA's actual blob throughput is—I checked the explorer twice to be sure—14 MB per day. That's it. That's the entire data availability demand for a protocol with a multi-billion dollar valuation.
I ran a back-of-the-envelope calculation. EigenDA's cost structure, if it were operated as a normal business, would require roughly $40,000 per month in server costs alone. It processes 420 MB per month. That's a unit cost of $95 per MB. My AWS bill for storing the same amount of data would be $0.01. The yield was sweet, but the exit was sharper.
The Ethereum Blob Trap
EIP-4844 was supposed to create a market for blob space. The idea: rollups would bid for blob inclusion, creating a competitive fee market that would make DA pricing efficient.
What actually happened? Blob utilization on Ethereum hovered around 3%. The blob fee market—the mechanism that was supposed to price DA correctly—collapsed. Blobs are free. Rollups don't need to compete for them. And here's the part that the modular DA maximalists don't want to talk about: if blobs are free and plentiful, the demand for a separate DA layer evaporates.
Why would a rollup pay Celestia or EigenDA when Ethereum is giving away blob space? The only answer the modular camp has is "sovereignty" and "customizability." But I've audited enough rollup stacks to know that 99% of teams want the path of least resistance. They want to deploy, get users, and not think about DA. Ethereum blobs are the path of least resistance.
The Rollup User Base Illusion
I want to zoom out for a second, because the DA utilization problem is actually downstream of a bigger issue: most rollups don't have meaningful usage.
I pulled DAU data for the top 20 rollups by TVL—Arbitrum, Optimism, Base, zkSync, Starknet, the whole lineup. Outside of the top five, average DAU is under 10,000. That's not a settlement layer problem. That's a product-market fit problem. No DA solution—no matter how cheap—solves for the fact that nobody is using these chains.
Now, I've been on the testing side of this before. In 2025, I signed up for several AI-agent-driven DeFi protocols to test their oracle feeds. What I found was a graveyard of abandoned integrations and half-baked dashboards. The DA layer situation is the same pattern: infrastructure was built for demand that hasn't materialized.
Contrarian: The DA Narrative Is a VC Exit Vehicle
Here's the angle that nobody's publishing. The modular DA narrative wasn't built because rollups needed DA. It was built because VCs needed a new sector to deploy into after the L1 wars exhausted themselves and the DeFi summer faded.
Liquidity fragmentation isn't a real problem—it's a manufactured narrative VCs use to push new products. The same playbook applies to DA. If you can convince the market that data availability is a critical infrastructure layer, you can raise a fund, seed a dozen "modular DA" startups, and exit into a token launch. The actual utility of the DA layer is secondary to the financial engineering.
I've counted the capital flows. Celestia raised $55M in a Series A and another $100M+ through its token sale. EigenDA is part of EigenLayer's broader ecosystem that has attracted billions in restaked ETH. Avail raised $43M. The aggregate capital deployed into dedicated DA layers exceeds $1B. The aggregate revenue generated by these layers? I calculated it at under $500,000 total, and most of that is from token incentives paid to the DA layers themselves.
That's a 2000x capital-to-revenue ratio. For comparison, even the most speculative DeFi protocols in 2020 had better fundamentals than that.
The other contrarian point: the DA layer "problem" was always overstated. Ethereum's blobs, even with low utilization, provide enough DA bandwidth for the entire rollup ecosystem for the next three to five years. The bottleneck isn't DA capacity. It's rollup execution and user adoption.
I've said this before and I'll say it again: the Data Availability layer is overhyped; 99% of rollups don't generate enough data to need dedicated DA. The rollups that do generate meaningful data—the top five by volume—are already posting to Ethereum blobs because it's cheaper and more secure. The modular DA thesis is a solution in search of a problem.
And here's the kicker: the intent-based architecture that many of these modular DA projects are pivoting to? It doesn't fix the DA problem. It just moves the MEV extraction from On-chain to off-chain solver networks. The DA layer becomes a compliance checkbox, not a value capture mechanism.
Takeaway: What to Watch When the Music Stops
So where does this leave us?
The DA layer sector is facing a fundamental demand problem. The tokens are down 60-80% from their highs, and the unlock schedules are going to keep pressure on prices. The projects with real technology—Celestia, EigenDA, Avail—will survive because they have the balance sheets to wait out the bear market. But survival isn't the same as relevance.
The signal to watch is straightforward: blob utilization on Ethereum is the real benchmark for DA demand. If Ethereum blobs stay under 10% utilization, there's no room for a dedicated DA layer to capture meaningful value. If they start hitting 50%+—which won't happen until the rollup ecosystem generates 10x more transactions than it does today—then the modular DA thesis might get a second look.
Until then, I'm watching the unlock schedules and the quiet exit of DA "partnerships" from rollup roadmaps. When a project stops mentioning its DA layer in public communications, you know the narrative is dead.
Listen to the whispers, but trust the ledger. The ledger says the DA layer is a ghost town. The whispers say it's the future. I know which one I'm trading.
The next 90 days will decide whether this sector consolidates into one or two survivors—or whether it becomes a cautionary tale about infrastructure built ahead of demand. I've got my order book open. Chaos is just data waiting for a pattern.