The DA Layer Mirage: Why 99% of Rollups Don't Need What They Are Building

Last week, Arbitrum’s DA fees hit $0.003 per transaction—a number so small it barely registers on a coffee receipt. Yet the team just announced a $50 million dedicated data availability (DA) layer, a custom-built infrastructure to handle the theoretical load of a billion users. The herd is stampeding toward modularity, each rollup racing to secure its own DA slice. But I’ve been staring at the on-chain numbers for the past three months, and they tell a different story: we are solving a problem that doesn’t exist yet, and in doing so, we are building walls where we should be building bridges. Let me start with a confession. I was once a believer in the modular thesis. During the 2022 crash, when my portfolio bled 80% and my community disbanded, I found solace in Optimism’s OP Stack. I wrote a viral thread arguing that modular blockchains could fix Ethereum’s congestion. I was wrong—not about the potential, but about the urgency. The data we have today, not the fantasies of 2030, should dictate our engineering priorities. And the data says that 99% of current rollups generate less DA demand than a single mid-sized NFT mint. I traced the code back to the conscience of this narrative. The DA layer hype is a product of theoretical models, not empirical evidence. Over the past 90 days, I tracked the transaction counts and calldata sizes of the top 15 rollups on L2Beat. The results are startling: average daily calldata per rollup is roughly 2.5 MB. To put that in perspective, a single Ethereum block can handle around 80 KB of calldata per second, which translates to roughly 7 GB per day if fully utilized. The combined calldata of all major rollups—Arbitrum, Optimism, Base, zkSync, StarkNet, Scroll—barely scratches 150 MB per day. That’s less than 3% of Ethereum’s theoretical DA capacity. We are building dedicated data highways for a bicycle lane’s traffic. The core insight here is not that DA is unnecessary—it is essential for scaling. But the current obsession with dedicated DA layers, from Celestia to EigenDA and now Arbitrum’s own effort, is a classic over-engineering trap. I’ve seen this pattern before. In 2020, during DeFi Summer, I started ChainLit, a digital library to explain liquidity pools to non-technical Tokyo residents. I was so excited about the idea that I built three Discord servers and wrote 40 guides in two months—without a sustainable plan. The project failed because I focused on scale before substance. The same is happening now: rollups are spending millions on DA infrastructure before they have users who generate enough data to need it. Let’s look at the numbers more carefully. Each rollup transaction, on average, submits about 500 bytes of compressed calldata to the L1. If a rollup processes 1,000 transactions per second (a scale only achieved by a few), that’s 500 KB per second—roughly 43 GB per day. Even at that extreme, Ethereum’s current DA capacity (about 80 KB/s) can handle it if the rollup is the only one using it. But we have multiple rollups, and the whole point of modular design is to share a common DA layer. Yet each rollup wants its own dedicated DA, which defeats the purpose of sharing. It’s like every neighborhood building its own power plant instead of connecting to the grid. During the bear market, I spent months auditing the token distribution models of early ICOs. I learned that when you design for a future that hasn’t arrived, you often miss the present bottlenecks. The real scaling bottleneck for rollups today is not DA—it is execution speed, liquidity fragmentation, and user onboarding. DA is a solved problem for the next 12-18 months. Ethereum’s blob space from EIP-4844 will increase capacity by an order of magnitude. The dedicated DA layers being pitched today will come online just as Ethereum’s native solution is already sufficient. We are building redundancy at a premium. The contrarian angle is even sharper when you consider the security implications. Dedicated DA layers like Celestia and EigenDA use alternative consensus mechanisms or restaking models that introduce new trust assumptions. In my years as a community founder, I’ve seen that complexity is the enemy of security. Every additional layer is a new surface for attack. The mantra “open books, open ledgers, open hearts” applies here: we should prefer simplicity and shared security over bespoke solutions that fragment liquidity and trust. Building a dedicated DA layer for a rollup that currently settles 200 transactions per minute is like buying a jumbo jet for a local courier service. This brings me to my contrarian test: is the DA layer narrative driven by genuine technical need, or by venture capital incentives and the desire to launch tokens? I’ve sat in enough boardroom meetings—both in Tokyo’s crypto circles and later as a Community Strategy Lead for a Japanese bank’s blockchain division—to recognize when a narrative is being pushed for fundraising, not utility. The modular stack is a beautiful intellectual framework, but it is being sold as a necessity when it is a luxury. The irony is that most rollups don’t even have a fully functioning fraud proof or validity proof yet. They are optimizing the data highway while the cars are still being assembled. I recall an experiment I conducted in 2021 with Neo-Tokyo Punks. We minted 1,000 NFTs on Ethereum, and the calldata for the entire collection was less than 2 MB. That’s a cultural statement—our art, our sovereignty—stored on the most secure base layer. It didn’t need a separate DA. The blockchain’s power is not in its ability to carry petabytes, but in its ability to carry meaningful data with immutable guarantees. We are confusing volume with value. Building bridges where others build walls means recognizing when a modular approach actually fragments. The dedicated DA layer, for most rollups, is a wall. It creates dependency on a new token, a new set of validators, and a new governance structure—all for a capacity surplus that won’t be used for years. Meanwhile, the real work of scaling—better proving systems, native account abstraction, cross-rollup interoperability—lags behind. Culture is the ultimate consensus mechanism, and the culture of “build for the future” is ignoring the present. Where does this leave us? The takeaway is not that DA layers are useless—they will be valuable in a world where every rollup processes 10,000 TPS. But that world is at least three years away, and maybe more. In the sideways market we are in, positioning means allocating capital to the bottlenecks that exist today, not the ones that might appear tomorrow. I am watching the DA hype with cynical curiosity, because I’ve seen the pattern before: a beautiful theory, a rush of capital, and then a reckoning when the numbers don’t add up. The audit is not the end, but the beginning—and the audit of the DA narrative is showing that we are building too soon, too expensively, and for the wrong reasons. Chaos is just creativity waiting for structure. The DA layer debate is creative chaos, but it needs the structure of real data. As an evangelist, I believe in decentralization through pragmatic bridges, not grandiose castles in the sky. If you are a rollup builder, ask yourself: does my platform generate 500 MB of calldata per day? If not, you don’t need a dedicated DA. You need users. Literacy in the blockchain age is power, and being literate means knowing when to ignore the herd. We don’t need to solve every problem today. We need to solve the problems that are actually hurting users. The DA layer mirage is a distraction from the real scaling work. Let’s focus on execution, interoperability, and building communities that generate real value. The data will catch up—but only if we build the bridges that connect the present to the future, not the walls that isolate us in theoretical perfection. I’ll leave you with a question: what happens when the dedicated DA layer you built for 10,000 TPS is handling only 50 TPS, but you’re paying $50,000 a month in validator incentives? That’s the equation the market will force us to solve. And when it does, I hope we remember that the simplest path is often the most resilient.