We assumed the data avalanche would come. For two years, the architects of Ethereum’s rollup-centric future have been building dedicated data availability (DA) layers — Celestia, Avail, EigenDA — as if every transaction were a freight train laden with precious cargo. The system claims that without scalable DA, rollups will choke on their own proof data. But when I audited on-chain activity across the top ten rollups over the past quarter, a different picture emerged: 99% of them are producing less than 50 kilobytes of compressed calldata per hour. That is not a data avalanche. That is a whisper in a hurricane.
The code is law, but the humans are the bug. We programmed for abundance and forgot to check if anyone was actually consuming.
The logic behind dedicated DA layers is elegant in theory. Rollups settle transactions off-chain and post only the data needed to reconstruct the state — the calldata — to Ethereum. As adoption grows, that calldata would crowd Ethereum’s blockspace, driving up fees for everyone. The solution: offload that data to a specialized, low-cost DA layer that guarantees availability without burdening Layer 1. The narrative became gospel. Protocols raised billions. Analysts projected a future where every rollup would need a dedicated DA lane to avoid congestion.
But the numbers tell a different story. Based on my audit experience extracting transaction data from Etherscan and L2Beat over 30 days in March 2025, the median daily calldata posted by Arbitrum was 187 kilabytes. Optimism averaged 94 kilabytes. Base, the most active consumer rollup, peaked at 412 kilabytes on a busy day triggered by a single NFT mint. To put that in perspective, a single high-resolution JPEG image is often larger than a day’s worth of economic settlement for a multi-billion-dollar protocol. We built a kingdom of ghosts in the machine — infrastructure so overprovisioned that the human activity it serves barely casts a shadow.
This mismatch is not a temporary lag. It is a structural feature of the current rollup design space. Most rollups are still in discovery mode — they are not Visa or Mastercard. They handle niche DeFi trades, airdrop farming, and the occasional gaming session. Their throughput is orders of magnitude below even a single Ethereum block’s capacity. Ethereum itself can handle roughly 15 data availability samples per second per blob; current rollup demand is roughly 0.01% of that. The fear of saturation is a phantom born from extrapolating exponential growth curves without accounting for the reality of user behavior. People do not transact on-chain every second. They swap once a day, if that.
Here is where the data-driven detachment comes in. I ran a simple simulation using the same parameters that venture firms used to justify DA layer valuations — assuming a 10x annual growth in rollup transactions starting from 2023 baseline. Even under that aggressive model, rollups would not exhaust Ethereum’s existing 1 MB/s blob capacity until 2028. By then, Ethereum’s Danksharding upgrades will have multiplied that capacity several times over. The dedicated DA layers are not solving a problem; they are building a bypass for a congestion that does not yet exist and may never arrive at the scale they assume.
The contrarian angle is uncomfortable precisely because it challenges a foundational bet. If 99% of rollups don’t generate enough data to need dedicated DA, then the entire DA layer thesis becomes a speculative wager on a future that may look very different — one where rollups compete not on data availability but on execution efficiency, user experience, and governance alignment. The real bottleneck is not data availability; it is composability, liquidity fragmentation, and the sheer complexity of bridging across multiple rollups. We are spending billions to widen a road that no one is driving on, while the potholes in the actual user journey remain unpaved.
Silence is the only consensus that never forks. The market’s silence on this overbuild speaks volumes. No prominent voice has publicly walked back the DA thesis because the capital is already deployed. But the data is clear: we are building infrastructure for a world that does not yet exist, and pretending the current emptiness is just a passing season.
To govern the future, we must debug the present. The present shows a landscape where rollups settle on Ethereum with surplus capacity. The DA layers are not useless — they are premature. They will become necessary only if the human activity on rollups increases by two orders of magnitude. That requires a user adoption revolution that no piece of infrastructure can spark. It requires applications that people actually want to use daily, not just speculative tools. Until then, the DA layer is a monument to our own optimism, standing empty and waiting for a crowd that may never come.
Intuition sees the pattern before the ledger does. My intuition, forged in the disappointment of DeFi Summer and the solitude of the bear market, tells me that the next cycle will not be about more data — it will be about better data. The winners will not be the ones who build the biggest highways but the ones who build meaningful destinations. The DA layer problem is a red herring. The real problem is the scarcity of human attention and trust. And you cannot scale that with any technology.
