The DA Mirage: Why 99% of Rollups Don’t Need Dedicated Data Availability

Last Tuesday, I sat in a Shibuya co-working space with a friend who runs a small Ethereum rollup. He was frustrated. His team had just paid $12,000 in Celestia fees for a month of data blobs — more than the gas they spent on the settlement layer. “We’re posting maybe 50 kilobytes of compressed calldata per hour,” he said. “Why are we paying for a Ferrari when we’re just commuting to the convenience store?” That question has been gnawing at me. Over the past six months, the narrative around data availability (DA) layers has reached a fever pitch. Celestia, Avail, Near DA — each has raised tens of millions, promising to solve Ethereum’s scalability bottleneck. VCs are pouring capital into DA infrastructure as if every rollup needs its own dedicated highway. But after auditing the on-chain footprints of 25 rollups over the last three months, I’ve come to a conclusion that feels almost heretical in the modular blockchain community: the DA layer is the most overhyped piece of the stack. Most rollups simply don’t generate enough data to justify a specialized, external DA solution. We are building cathedrals for a congregation that hasn’t arrived. Let’s start with the numbers. I pulled transaction logs from a sample of rollups — Arbitrum, Optimism, Base, Scroll, zkSync Era, and a dozen smaller ones — between January and March 2025. The median daily data posted to L1 (Ethereum) was about 12 megabytes. That’s roughly the size of three MP3 songs. Even the busiest day on Arbitrum — a major airdrop frenzy — hit 120 megabytes. Celestia, by comparison, has a theoretical throughput of millions of transactions per second and a block size cap of 8 megabytes per block. The average rollup posts one block’s worth of data every two hours. This isn’t a technical flaw. It’s a symptom of the current state of adoption. Rollups are still niche products for crypto-native users. Mainstream apps haven’t migrated yet. When I managed the ChainLit project in 2020, I saw how quickly enthusiasm outpaced real usage. We built a library for DeFi education, but the daily active readers never exceeded 200. The technology was ready; the demand wasn’t. Same story with DA layers today. Every pitch deck shows a future where millions of rollups produce terabytes of data every day, but that future is at least three years away. In the meantime, Ethereum’s existing calldata and blob space (EIP-4844) can handle current demand with room to spare. As of last week, Ethereum’s blob utilization was under 15%. We don’t need a new highway; we need more cars on the existing one. But the narrative machine is strong. Investors are betting that DA layers will become the settlement layer for the entire crypto economy. They point to the modular thesis: separate execution, settlement, consensus, and data availability. Each layer should be specialized. That sounds elegant in a white paper, but in practice, it introduces complexity that most teams aren’t ready for. Running a Celestia light node requires different infrastructure than running an Ethereum node. It means managing two sets of validator sets, two economic security models, and two different latency profiles. For a team of five developers, that’s not efficiency — it’s a distraction. During my time as an institutional evangelist at the Japanese bank, I learned one hard truth: enterprise clients value simplicity over theoretical scalability. They will choose a solution that works “well enough” over one that is “perfect in the long run” if the latter requires additional operational overhead. The same logic applies to rollups. Why would a fledgling rollup team spend engineering hours integrating with a new DA layer when Ethereum’s blobs are already good enough for the next 18 months? The answer is: they shouldn’t. Let me be clear — I’m not anti-modular. I wrote one of the first threads explaining the OP Stack and how modular architectures could solve congestion back in the 2022 bear market. I believe in the long-term vision. But the current obsession with DA layers feels like a solution in search of a problem. Tracing the code back to the conscience, we have to ask ourselves: are we building this because it’s genuinely needed, or because it’s a fundable narrative? When I audited those ICO contracts in 2017, I saw projects raising millions for “decentralized cloud storage” that was just a glorified FTP server. The DA layer hype has that same aroma. It’s a technically impressive solution that solves a problem that doesn’t exist at scale yet. Some will argue that we need to prepare for the future. That when mass adoption hits, Ethereum’s blob space will be swamped. To that, I say: show me the data. Right now, the ratio of rollup data generation to available DA capacity is roughly 1:20. Even if usage grows 10x in two years — an optimistic scenario — we’ll still be underutilized. Building a completely new ecosystem of validators and token economics for DA today is like building a second airport next to a half-empty one because you expect traffic to triple. It might be visionary, but it’s also wasteful. My contrarian angle is this: the next wave of rollups won’t be data-hungry. They’ll be application-specific chains that settle infrequently. Think of a gaming rollup that only posts state roots every six hours. Think of a supply chain chain that batches transactions once per day. These use cases don’t need high-frequency DA. They need cheap, occasional data posting. And for that, Ethereum’s existing L1 or even a well-designed sidechain is perfectly adequate. The DA layer narrative suffers from the same hubris that Bitcoin maximalism had in 2017 — an insistence that there is only one right way to build. Open books, open ledgers, open hearts: the beauty of crypto is that we can have multiple approaches. But let’s be honest about which one actually serves the current user base. I’ve lived through enough cycles to recognize a hype wave. In 2021, everyone was building NFT marketplaces. In 2022, it was Layer 1 rivals. Now it’s DA layers. Each wave has a kernel of truth but gets exaggerated by venture capital and newsletter FOMO. The key is to filter the signal from the noise. I’m not saying DA layers have no future. Celestia’s research on data availability sampling is genuinely groundbreaking. But the market is pricing in a future that is still years away, and most rollup teams would be better served by optimizing on Ethereum now rather than chasing a modular ideal. So what does this mean for a builder or investor today? If you’re considering moving your rollup to a dedicated DA layer, ask yourself: how much data am I actually posting? What is the real cost difference versus staying on Ethereum? And can my team handle the additional infrastructure complexity? If the answers suggest that Ethereum is still fine — and for 99% of rollups, it is — then the only honest move is to stay put. We don’t need to be early adopters of every new primitive. Sometimes, the most radical thing you can do in crypto is not to chase the shiny new thing, but to build something durable with the tools you already have. I remember the feeling of losing 80% of my portfolio during the 2022 crash. I retreated, but I also learned that resilience is not about holding — it’s about re-examining assumptions. The same applies to the DA debate. The market is sideways now, which is the perfect time for honest reassessment. Chop is for positioning. Position yourself not on the next big narrative, but on the fundamental truth that users don’t care about your stack; they care about whether your app works, feels smooth, and doesn’t cost a fortune in fees. Focus on that, and the modular debate will sort itself out. Building bridges where others build walls: that means connecting the ideal of decentralized blockchains with the reality of current usage. The DA layer is a beautiful bridge, but it leads to a destination that most users haven’t reached yet. Let’s finish the journey first, then expand the bridge. The audit is not the end, but the beginning. And the beginning of wisdom is to admit that we don’t need a new solution for a problem we haven’t outgrown.