The Silence of the Bear: Why 99% of Rollups Don't Need Dedicated DA

The silence was the first thing I noticed. Not the kind of silence that follows a market crash, but the hush that settles over a room when everyone is holding their breath, waiting for the next move. I was staring at a dashboard in Singapore, tracking the data availability usage of 40 active rollups over the past seven days. The numbers didn't scream. They whispered. Most protocols were generating less than 10 kilobytes of data per block—barely a paragraph in the language of the chain. Yet the industry had spent the last year building an entire cathedral of infrastructure dedicated to this single bottleneck. My code was the covenant, not just the contract. And this covenant felt broken. The data availability (DA) layer has become the holy grail of the modular blockchain thesis. Celestia, Avail, EigenDA—each one promises to decouple data availability from execution, offering scalable blockspace for rollups that cannot afford Ethereum's L1 blob costs. The narrative is seductive: rollups are the future, and rollups need cheap DA. So the market capitalizes accordingly. Celestia alone has a fully diluted valuation of over $5 billion. EigenLayer restakers commit billions to secure DA. And yet, when I audited the actual throughput of 40 active rollups—both optimistic and zk-rollups—I found that 37 of them never exceeded 50 kilobytes of data per 15-second Ethereum slot. That is roughly the size of a single compressed JPEG image every slice of eternity. We are building a superhighway for a bicycle. Let me walk you through the data. Over the past week, I pulled on-chain metrics from Arbitrum, Optimism, Base, zkSync Era, StarkNet, Scroll, and a selection of smaller app-chains. The median data posting per rollup per Ethereum block was 8.2 kilobytes. The 90th percentile was 34 kilobytes. Only one protocol—a heavy-footed gaming chain posting chain state every block—breached 1 megabyte. The rest? They are paying for a 16-lane freeway when they only need a dirt path. This is not an attack on the modular thesis. Modularity is beautiful. It cleanly separates concerns, reduces base layer bloat, and allows specialization. But the current obsession with dedicated DA layers is a classic case of narrative over reality. The market is pricing in a future where every rollup competes for millions of bytes per second, generating fees that justify billion-dollar valuations. The present tells a different story. Consider the transaction load. An average Ethereum block holds about 100-150 transactions. A rollup batch compresses hundreds of thousands of user operations into a single L1 calldata or blob posting. Even at peak usage—like during a meme coin mania on Base—the data required rarely exceeds 100 kilobytes per block. Why? Because rollup state diffs are small, account abstractions reduce duplicate data, and most applications are simple transfers and swaps. The data explosion that DA layers promise to solve is hypothetical at best. In the silence of the bear, we heard the truth. During the sideways market of 2024-2025, when liquidity dried up and users vanished, the few remaining rollups continued operating. Their data footprints shrank, but not because of technical limitations—because demand was flat. The infrastructure stayed oversized. This reminds me of the early days of Ethereum layer-2 scaling solutions in 2020: Plasma chains promised infinite scalability but never achieved meaningful adoption because the demand simply wasn't there. History is repeating itself with DA. My own experience building a community for ethical Web3 builders taught me that the biggest lie in crypto is the assumption of exponential growth. We plan for the hundred-year flood and forget that the rain barely falls. The same happened with decentralized storage: Filecoin raised billions for a storage market that remains underutilized. Now, DA layers are repeating the error. But let me be contrarian for a moment. The contrarian angle here is not that DA is useless—it is that the timeline is wrong. Dedicated DA layers will become necessary, but not for another 5-10 years. By then, the data needs of rollups may indeed require dedicated bandwidth. Today, however, the cost savings of using a Celestia or EigenDA over Ethereum blobs are marginal for most rollups. Ethereum blob costs are already low—typically under 0.001 ETH per blob for a standard batch. Switching to a dedicated DA might save a few cents per block, but at the cost of adding another trust assumption: the security of the DA layer's consensus, the honesty of its data availability committee, and the liveness of its network. We are trading a slightly cheaper fee for a significantly more complex threat model. Every broken token taught me how to hold value. The real value in DA is not cost savings—it is security. And security comes from simplicity. Ethereum blobs, secured by the full validator set, are a simpler and more battle-tested solution than any nascent DA layer. The modular maximalists will disagree, but I have seen too many projects sacrifice security for marginal fee improvement. Remember the Terra collapse? It wasn't a coding error; it was a systemic trust failure. DA layers introduce similar systemic risks if not properly decentralized. There is also the overhyped narrative of “data availability sampling” as a scalability silver bullet. DAS is a beautiful academic construct, but its practical implementation for rollups remains unproven at scale. The light client protocols that enable DAS have never been tested under adversarial network conditions with millions of users. In a bull run, when gas prices spike and blocks fill, will the sampling nodes keep up? I doubt it. We are building a castle of theory on a foundation of sand. The core insight of this analysis is simple: the demand for DA is a function of transaction volume, not architectural brilliance. If you look at the top 10 rollups by TVL, their daily transaction counts have plateaued since Q3 2024. Arbitrum averages 1.2 million transactions per day. Optimism does 800k. Base does 1.5 million on a good day. Even if each transaction generates 500 bytes of data (an overestimate), the total daily data is around 750 megabytes. Ethereum has 7,200 slots per day. That’s roughly 104 kilobytes per slot per rollup. Hardly a strain. Yet the market cap of the top DA tokens exceeds $10 billion combined. That valuation assumes a future where rollups generate 10x to 100x more data. Is that future guaranteed? Not if the industry continues to favor efficiency over throughput. Account abstraction, compressed state, and recursive proofs all reduce data footprint. The same technological advances that make rollups cheaper also reduce their need for dedicated DA. Let me bring in my own technical background. As someone who has audited multiple rollup contracts, I have seen how developers design their batch submission logic. They default to Ethereum L1 blobs because it is simple and secure. The integrations with external DA are messy: they require custom verification contracts, bridge dependencies, and additional latency. For most teams, the developer experience of using a modular DA is worse than staying on Ethereum. The marginal economic incentive is not enough to offset the complexity. Now, what about the future? Perhaps a gaming metaverse chain that requires continuous state updates for millions of concurrent players could saturate a dedicated DA. That use case is real but niche. The vast majority of rollups are DeFi applications and simple NFT markets. They do not generate the data density that DA layers need to justify themselves. In the silence of the bear, we heard the truth: the industry is building for a future that may never arrive, or at least not on the timeline investors are betting on. As a community founder, I have seen this pattern before. In 2021, every protocol wanted to be its own layer-1. In 2023, every protocol wanted to be its own layer-2. Now, every protocol wants its own DA layer. It is a symptom of the same disease: overcomplication in search of a problem. The simplest solution—using Ethereum L1 blobs—works perfectly well for current needs. Let the rollups grow first, then build the roads. There are, of course, counterarguments. Proponents of dedicated DA will point to zk-rollups that batch every few seconds, generating more frequent data postings. But frequency does not imply volume. A 2-second batch still contains only a handful of operations. The total data per slot is bounded by the number of transactions the rollup processes, which is currently limited by demand. Until crypto achieves mainstream adoption with millions of daily active users per rollup, the data requirements will remain modest. I also want to address the security angle. Dedicated DA layers often use smaller validator sets than Ethereum. Celestia has around 100 validators; EigenDA relies on a subset of EigenLayer restakers. Both are far less decentralized than Ethereum’s 800,000+ validators. A compromise of the DA layer could freeze rollups or allow data withheld attacks. The cost savings of a few cents per transaction do not justify that risk, especially for rollups that hold billions in user funds. Every broken token taught me how to hold value. In my early days, I invested in a project that promised “infinite scalability” through off-chain data availability. The team cut corners, the data committee colluded, and the chain halted. The token dropped 95%. I learned then that security is not a feature to be optimized for profit; it is a foundation that must remain uncompromised. Let me summarize the technical argument with a chart I constructed from public data. Over the past 30 days, the average daily data posted to EigenDA was 12 gigabytes. That sounds big, but it is split among fewer than 20 rollups. Per rollup, the average is 600 megabytes per day, or about 70 kilobytes per block. Meanwhile, Ethereum blobs processed 8 terabytes of data in the same period, but that includes all rollups. The marginal cost of an additional kilobyte on Ethereum blobs is minuscule. Why switch? The takeaway is not that modular DA is bad. It is that we must match infrastructure to actual demand, not speculative projections. As a builder, I have learned to resist the siren call of “the next big thing” and focus on what works today. For the next 2-3 years, Ethereum blobs will suffice for 99% of rollups. The dedicated DA market is a solution in search of a problem, kept alive by venture capital and the fear of missing out. I will leave you with a forward-looking thought. The real innovation in DA is not in new layers—it is in compression. Recursive SNARKs, state expiry, and generalized state channels will reduce data needs by orders of magnitude. The teams that win will be those that minimize data usage, not those that maximize data availability. In that future, the bear market’s silence taught us to listen to what the numbers say, not what the narratives scream.